Contribution à l organisation des opérations d escale dans une plateforme aéroportuaire

Transcription

1 Contrbuton à l organsaton des opératons d escale dans une plateforme aéroportuare Salma Ftour-Trabels To cte ths verson: Salma Ftour-Trabels. Contrbuton à l organsaton des opératons d escale dans une plateforme aéroportuare. Automatc. Insttut Natonal Polytechnque de Toulouse - INPT, 203. Englsh. <tel > HAL Id: tel Submtted on 8 Apr 204 HAL s a mult-dscplnary open access archve for the depost and dssemnaton of scentfc research documents, whether they are publshed or not. The documents may come from teachng and research nsttutons n France or abroad, or from publc or prvate research centers. L archve ouverte plurdscplnare HAL, est destnée au dépôt et à la dffuson de documents scentfques de nveau recherche, publés ou non, émanant des établssements d ensegnement et de recherche franças ou étrangers, des laboratores publcs ou prvés.

2 En vue de l'obtenton du DOCTORAT DE L'UNIVERSITÉ DE TOULOUSE Délvré par : Insttut Natonal Polytechnque de Toulouse (INP Toulouse) Dscplne ou spécalté : Systèmes Industrels Présentée et soutenue par : Mme SALMA FITOURI TRABELSI le mard 26 novembre 203 Ttre : CONTRIBUTION A L'ORGANISATION DES OPERATIONS D'ESCALE DANS UNE PLATEFORME AEROPORTUAIRE. Ecole doctorale : Systèmes (Systèmes) Unté de recherche : Laboratore de Mathématques Applquées, Informatque et Automatque pour l'aéren - MAIAA Drecteur(s) de Thèse : MME CATHERINE MANCEL M. FELIX MORA-CAMINO Rapporteurs : M. CARLOS ALBERTO NUNES CONSENZA, UNIVERSITE FEDERALE DE RIO DE JANEIRO M. WALID EL MOUDANI, UNIVERSITY EL KOBBEH Membre(s) du jury : M. PHILIPPE MARTHON, INP TOULOUSE, Présdent M. BATATIA HADJ, INP TOULOUSE, Membre M. FELIX MORA-CAMINO, LAAS TOULOUSE, Membre M. JOSÉ LUIS MOURA BERODIA, UNIVERSIDAD DE CANTABRIA, Membre Mme CATHERINE MANCEL, ECOLE NATIONALE DE L'AVIATION CIVILE, Membre M. SERIGNE GUEYE, UNIVERSITE D'AVIGNON, Membre

3 Acnowledgements Ths doctoral research was prepared wthn MAIAA laboratory of Ar Transport department at ENAC. I would le to express my deepest grattude to my thess supervsor Professor Félx Mora- Camno for hs contnuous gudance and support throughout the research. Hs encouragement and advce led me to the rght path and are greatly apprecated. I would le also to than all the members of my jury, especally Professor Carlos Alberto Nunes COSENZA from the Federal Unversty of Ro de Janero and Professor Wald EL MOUDANI from the Faculty of Busness of the Lebanese Unversty ther acceptance to revew my thess dssertaton. Also, I would le to than Professor Phlppe MARTHON for hs acceptance to char the jury of my PhD thess. My heartfelt apprecaton also goes to the frends and colleagues at the Automaton Research Group of MAIAA laboratory. I would also le to extend my deepest grattude to my famly for ther uncondtonal love and support. I am also grateful to everyone who, n one way or another, has helped me get through these years.

4

5 To my father s soul To my mother, to my sster To all my famly, to all my frend

6

7 Abstract The ncrease of the world ar traffc growth of the last decades has generated a permanent challenge for cvl avaton authortes, arlnes and arports to supply suffcent capacty to provde a safe transportaton servce wth acceptable qualty standards. New traffc management practces, such as A-CDM, based on mult-agent and collaboratve decson mang concepts have been ntroduced at arports. However, wthn the turnaround process of arcraft at arports, ground handlng management of arcraft has not been developed specfcally n the A-CDM approach, even f t has an mportant role n the fludty of arcraft operatons at arports. The man objectve of ths thess dssertaton s to contrbute to the organsaton of the ground handlng management at arports. It conssts to provde a structure organze the ground handlng management compatble wth the A -CDM concept. The proposed structure ntroduces a ground handlng coordnator (GHC) whch s consdered as an nterface for communcaton between the partners of the A -CDM and the dfferent ground handlng managers (GHM). Ths herarchcal structure allows sharng nformaton wth partners n the A -CDM on the one sde and on the other sde, nteractng wth ground handlng managers (GHM). Decson mang processes based on heurstcs have been developed at each level of the proposed organzaton and have been also evaluated n the case of nomnal condtons and n the case of the presence of major dsruptons. Key words: arport management, ground handlng operatons, CDM, mult-agent system

8

9 Résumé La crossance du trafc aéren a rendu crtque l opératon de la geston des plateformes aéroportuares. Celle-c fat appel à de nombreux acteurs (autortés aéroportuares, compagnes aérennes, contrôle du trafc aéren, prestatares de servces, ). Le concept d Arport Collaboratve Decson Mang (A-CDM) développé depus une dzane d années est basé sur un partage d nformatons opératonnelles en temps réel entre les dfférents acteurs de la plate-forme, permettant de prendre des décsons en commun pour rechercher une utlsaton optmale, en toutes condtons, des capactés de l aéroport. L objectf prncpal de cette thèse est de contrbuer à l organsaton de la geston des opératons d escale dans une plateforme aéroportuare. Il s agt de proposer une structure d organsaton de cette opératon qu sot compatble avec l approche A-CDM. La structure proposée ntrodut un coordnateur des opératons d escale (GHC) qu joue le rôle d nterface de communcaton entre les partenares de l A-CDM et les dfférents gestonnares des opératons d escale (GHM). Cette structure hérarchque permet d une part de partager des nformatons avec les partenares de l A-CDM et d autre part d nteragr avec les gestonnares des opératons d escale (GHM). Les processus de prse de décson basés sur des heurstques ont été développés à chaque nveau de l organsaton proposée et sont évalués auss ben dans le cas de condtons nomnales que dans le cas de la présence de perturbatons majeures. Mots clé : geston des aéroports, actvtés d assstance en escale, CDM, systèmes multagents

10

11 Contents Contents GENERAL INTRODUCTION.... CHAPTER THE GROUND HANDLING PROCESS AT AIRPORTS Introducton Identfcaton of ground handlng Poston of ground handlng n arport system operatons Detaled analyss of the man ground handlng processes The passenger boardng/de-boardng processes The luggage loadng/unloadng processes The cleanng process The caterng process The fuellng process Potable water supply and santaton process The de-cng process Push-bac Ground handlng as a complex mult-actvty process Examples of ground handlng processes The temporal dmenson of Ground Handlng Crtcal path analyss of ground handlng process Concluson CHAPTER 2 ANALYSIS OF THE ORGANIZATION OF GROUND HANDLING MANAGEMENT AT MAJOR AIRPORTS Introducton The ground handlng staeholders Arports, arlnes and ground handlng operators... 34

12 Contents The current stuaton wth respect to Ground Handlng The mportance of managng ground handlng Ground handlng costs Costs of Ground Delays for Arlnes Tme Scales for Ground Handlng Management Strategc plannng for ground handlng Tactcal plannng for ground handlng Operatonal plannng for ground handlng Real-tme management for ground handlng Concluson CHAPTER 3 Optmzaton Approaches for Ground Handlng Operatons: An Overvew Introducton Management of an arsde passenger bus fleet Problem defnton Problem class Problem formulaton Soluton approaches and comparatve results Management of fuellng trucs at arsde Problem defnton Problem classfcaton Mathematcal formulaton Soluton approach Acheved performances Management of a connectng baggage fleet Problem defnton Class of problem... 58

13 Contents Mathematcal formulaton Proposed soluton approach Obtaned results Management of a de-cng fleet Problem descrpton Current studes Mathematcal formulaton Soluton approach and results Management of caterng fleets Problem descrpton Mathematcal formulaton Global approaches A Centralzed Approach for the Ground Handlng Assgnment Problem Decentralzed Approaches of the Global Ground Handlng Assgnment Problem Analyss and concluson CHAPTER 4 A GLOBAL ORGANIZATION OF GROUND HANDLING MANAGEMENT Introducton A-CDM and ground handlng management The A-CDM concept Operatonal prncples of CDM Ground handlng and CDM Introducng an Arport Ground Handlng Coordnator Ground handlng mlestones montorng by GHC Ground Handlng Coordnaton Global plannng of ground handlng resources... 88

14 Contents 4.5. Decentralzed ground handlng management Local ground handlng management Coordnaton level of ground handlng management Petr Net representaton of proposed ground handlng organzaton and operaton Ground handlng unts Ground handlng manager Ground handlng coordnator Concluson CHAPTER 5 DECISION MAKING PROCESSES FOR THE PROPOSED GLOBAL APPROACH Introducton The Central Planner Problem Adopted notatons Tentatve problem formulaton Analyss and soluton process Numercal applcaton The proposed GHC heurstc Decentralzed fleet management Classes of fleet management problems Adopted notatons Formulaton of the GHFAS problem (C case) Formulaton of the GHFAS problem (C2 case) On lne Ground Handlng Fleet Assgnment (GHFA) problem at the level of each GHM Ground Handlng Fleet Coordnaton Proposed heurstcs for on-lne GHFA Case study Arport and ground handlng characterstcs... 2

15 Contents Implementng the herarchcal approach Implementng the global plannng of ground handlng resources Implementng the heurstcs for on-lne GHFA Concluson CHAPTER 6 GROUND HANDLING MANAGEMENT UNDER DISRUPTION Introducton Arport Dsrupton Defnton of arport dsrupton Consequences of arport dsrupton Sources of arport dsrupton Ground Handlng Management Objectves and Operaton under Arport Dsrupton Ground handlng management objectves under arport dsrupton A proposal for ground handlng management under arport dsrupton Operatonal uncertanty durng arport dsrupton Some elements about fuzzy dual numbers Fuzzy dual delays and duratons Ranng Flght under Dsrupton wth Uncertanty Ground Handlng Fleets assgnment to flghts Illustraton of the proposed approach Concluson CONCLUSION AND PERSPECTIVES... 5 BIBLIOGRAPHIES ANNEX I ANNEX II ANNEX III... 8 ANNEX IV ANNEX V... 99

16 Contents ANNEX VI

17 Lst of Fgures Lst of Fgures CHAPTER Fgure. : Localzaton of ground handlng wthn the turnaround process... Fgure. 2: Arcraft related operatons at arports... 2 Fgure. 3: Dfferent devces to handle passengers boardng and de-boardng processes... 5 Fgure. 4: Luggage loadng/unloadng processes... 5 Fgure. 5: Luggage loadng/unloadng processes... 6 Fgure. 6: Caterng process... 7 Fgure. 7: Dfferent arcraft fuellng processes... 8 Fgure. 8: Santaton process... 8 Fgure. 9: Potable water supply process... 9 Fgure. 0: On-gong de-cng process Fgure. :Push- bac process Fgure. 2: Arcraft servcng arrangement Typcal handlng operatons Boeng ER [Boeng manual]... 2 Fgure. 3: Ground handlng process for a Boeng B737 [Boeng 737 Manuel] Fgure. 4: Ground handlng process at Belgrad Internatonal Arport [Vdosavljevć and al, 200].22 Fgure. 5: Ground handlng process at Stocholm Arport [Norn and al 2008] Fgure. 6: Turnaround wth loose ground handlng actvtes Fgure. 7: Turnaround wth tght ground handlng actvtes Fgure. 8: Typcal duratons of handlng operatons Boeng [Boeng B777 manual] Fgure. 9: Typcal duratons of handlng operatons Arbus [Arbus A330 manual] Fgure. 20: Canddate crtcal paths for ground handlng process CHAPTER 2 Fgure 2. : Turnaround charges for an Arbus 320 at dfferent arports 203 [Zurch Arport, 203]... 4 Fgure 2. 2: Management tmelne... 46

18 Lst of Fgures CHAPTER 4 Fgure 4. :The arport partners nvolved n the A-CDM Fgure 4. 2 : Connecton of A-CDM wth Ground Handlng Fgure 4. 3 : Example of the set of ground handlng actvtes for an A320 at Stocholm arport Fgure 4. 4: Introducng a Ground Handlng CDM Fgure 4. 5 : Plannng of a safe level for ground handlng resources Fgure 4. 6 : Three-levels organzaton of ground handlng management Fgure 4. 7 : RdP representaton of GHU s operatons Fgure 4. 8 : RdP representaton of operatons by a GHMj Fgure 4. 9 : RdP representaton of operatons by a GHC... 0 CHAPTER 5 Fgure 5. 2 : Vehcles routes... 3 Fgure 5. : Structure and duraton of the ground handlng actvtes... 3 Fgure 5. 3 : 0 /08/2007 Palma de Mallorca Arport Arcraft hourly traffc... 2 Fgure 5. 4 : Hourly delays dstrbuton resultng from the proposed heurstc Fgure 5. 5 : Nomnal composton of ground handlng fleets Fgure 5. 6: Number of the resources requred for each ground handlng actvtes each of perod of tme Fgure 5. 7 : Hourly delays dstrbuton resultng from the proposed heurstc CHAPTER 6 Fgure 6. : Ground handlng management under dsrupton Fgure 6. 2 : Operatonal uncertanty durng arport dsrupton Fgure 6. 3 : Representaton of a fuzzy dual number Fgure 6. 4 : Examples of shapes fro fuzzy dual numbers Fgure 6. 5 : Example of nequaltes (wea and strong) between fuzzy dual numbers... 4 Fgure 6. 6 : Examples of fuzzy equalty between fuzzy dual numbers... 4 Fgure 6. 7 : Example of ground handlng actvtes sequencng Fgure 6. 8 : Illustraton of the duraton ~ j Fgure 6. 9 : The hourly dstrbuton of delayed arcraft at departure (Scenaro ) Fgure 6. 0 : The hourly dstrbuton of delayed arcraft at departure (Scenaro 2)... 48

19 Lst of tables Lst of Tables CHAPTER Table. : Scope of ground handlng operatons [Ashford and al. 203]... 3 Table. 2: Mnmal and maxmal value for the ground handlng process CHAPTER 2 Table2. : Dstrbuton of responsbltes for ground handlng operatons at 72 selected arports [Norman and al. 203] Table2. 2: Departure delay causes [Clare and al, 2004] CHAPTER 5 Table5. : The nomnal arrval and departure schedules and the parng postons of arcraft.. 3 Table5.2 : The assgnment soluton... 4 Table5. 3 : Soluton of herarchcal approach Table5. 4: The unt tme perod of each ground handlng operaton results CHAPTER 6 Table6. : Effectve arrvals and scheduled departures Table6. 2: Statstcal results for dsrupton scenaros... 47

20 Lst of tables

21 General Introducton GENERAL INTRODUCTION

22 2 General Introducton

23 General Introducton Along the last decades of worldwde growth of ar traffc, the ar transportaton system (ATS) has been developng new mproved operatonal procedures based on the up to date avalable nformaton processng technology. Ths started as early as 962, wth for example, the creaton of the AGIFORS (Arlnes Group of IFORS) by man arlnes usng the frst manframe computers avalable n that epoch. Today n the Internet era, the operatons of the Ar Transportaton System nvolve drectly global actors (arports, arlnes, ar traffc control (ATC), ar traffc management (ATM)) as well as local actors (ground handlers, local supplers ) through nterconnected nformaton networs. The management of arports plays an mportant role wthn ths complex system snce demand for ar transportaton s arport referenced (they are at the same tme orgn and destnaton for the flghts) and many effectveness ndexes are based on events occurrng at the arport and the correspondng statstcs. Besdes safety and securty whch are a prorty ssues and they provde the operatonal envronment at arports, arcraft traffc delays at arports and more partcularly flght departure delays, are a also seen as permanent ssues for arport management. Part from managng ar traffc delays, safety and securty, other man objectves of the traffc management at arports are the mprovements of operatonal effcency by reducng the arcraft delays, the optmzaton of arport resources to reduce costs and the ncreased predctablty of effectve flght departure tmes. In fact, for many years now, flght delays are one of the most mportant problems n the ar transportaton sector. For nstance, n % of all European flghts were late more than 5 mnutes at departure [Frce and al, 2009]. These recurrent delays resulted n a lower qualty of servce to passengers whle arlnes and arports were also affected wth a loss of effcency and consequently wth a loss of ncomes and whle the envronmental performance of the ATS s downgraded (ncreases fuel consumpton and emssons of partcles). If delays resultng from bad weather are mostly unavodable, delays resultng from nsuffcent performance of traffc management at arport may be reduced by searchng for new operatonal approaches ams at mprovng the overall arport performance. Arport Collaboratve Decson Mang (CDM) [Eurocontrol, 203] s a recent concept whch creates a common ground for the dfferent components of the ATS. Ths concept s based on an mproved communcaton between the dfferent actors of the arport (Ar Traffc Control, Arport Authortes, and Arlnes). CDM has already been appled to some major European arports where t has mproved ther performances and has receved a good acceptance by the 3

24 General Introducton dfferent actors. However, wthn the turnaround process of arcraft at arports, ground handlng management of arcraft has not been developed specfcally n the CDM approach, even f t has an mportant role n the fludty of the arcraft ground movements at arports. The man objectve of ths PhD thess s to contrbute to the development of an effcent management organzaton of ground handlng at arports whch should be compatble wth the CDM approach. Ground handlng addresses the many servces requred by a transportaton arcraft whle t s on the ground, pared at a termnal gate or a remote poston n an arport, ether at arrval from a last flght or at departure for a new flght. Ths ncludes the processng of boardng/de-boardng passengers, baggage and freght, as well as the arcraft tself (fuellng, cleanng, santaton, etc). Ths thess s organzed n sx man chapters, concluson and annexes. In Chapter, the general ground handlng process at the level of a partcular flght s dentfed and descrbed. Then each classcal ground handlng actvty s detaled. Fnally the tme dmenson of the ground handlng attached to a partcular flght s dscussed. In Chapter 2, the man manageral ssues wth respect to ground handlng management at the arport are consdered: ground handlng management organzaton wth the possble roles of the dfferent staeholders, ground handlng costs and beneft ssues and fnally the dfferent tme scales adopted for ground handlng management. In Chapter 3, an overvew of quanttatve approaches to solve ground handlng decson problems at the operatons level s performed. Specfc as well as global approaches mang use of classcal mathematcal programmng approaches or more recent computatonal approaches are consdered. In Chapter 4, a global organzaton of ground handlng management at arports, ncludng a ground handlng coordnator and compatble wth the CDM approach s developed, analyzed and dscussed. In Chapter 5, wthn the manageral framewor proposed n the prevous chapter, an heurstc based soluton approach of the man operatons problems encountered n ground handlng at arports s proposed. Then a case study s developed. 4

25 General Introducton In Chapter 6, also wthn the same manageral framewor, the case of arport dsrupton s treated at the ground handlng level. Fnally, the Concluson Chapter provdes a summary of the contrbutons of ths wor as well as the man perspectves for ts applcaton as well as subsequent developments n the same lne. The dfferent annexes provde some theoretcal and practcal bacground wth respect to the technques used n ths PhD report. 5

26 6 General Introducton

27 Chapter The Ground Handlng at Arport. CHAPTER 2. THE GROUND HANDLING PROCESS AT AIRPORTS 7

28 Chapter The Ground Handlng at Arport 8

29 Chapter The Ground Handlng at Arport.. Introducton Ths thess focuses on the ground handlng management at arports. From one arport to another, dependng on ther physcal desgn, composton of traffc and many other factors, ground handlng actvtes can appear to be performed very dfferently. So, to clarfy our feld of study, n the frst step of ths chapter, the concept of ground handlng adopted n ths thess s presented and dscussed. It appears then that even f some traffc management related actvtes and arlnes related crew and arcraft management ssues are not ncluded n ths concept, the ground handlng actvtes realzed on a grounded arcraft would result n a very complex process. Then, n the second step, n ths chapter, a detaled descrpton of the man ground handlng actvtes performed on a transportaton arcraft s proposed. These man actvtes cover: passenger de-boardng, passenger boardng, caterng, cleanng, fuellng, push-bac. Fnally the whole ground handlng process performed on a grounded arcraft s consdered through dfferent examples of smulaton whle ts tme dmenson s ntroduced and dscussed..2. Identfcaton of ground handlng Arcraft ground handlng s composed of a set of operatons appled to an arcraft to mae t ready for a new commercal flght or to fnalze an arrvng commercal flght. In general techncal and commercal crew actvtes at arrval and departure are performed by the arlnes and are not consdered to be part of the ground processng actvtes. It s the same wth the arcraft mantenance actvtes whch are realzed, n accordance wth regulatons, durng the stopover of the arcraft, n parallel wth the ground handlng actvtes. A typcal ground handlng process s composed of the followng steps: De-boardng passengers, unloadng baggage, fuellng, caterng, cleanng, santaton, potable water supply, boardng passengers, loadng baggage, de-cng and pushng bac the arcraft. Ground handlng actvtes can be processed at dfferent perod of tme and places n the arport. 9

30 Chapter The Ground Handlng at Arport Techncal and commercal crew de-board the arcraft once all passengers have left the arcraft whle other arrval ground handlng actvtes can be performed. Dependng of the turnaround characterstc (short turnaround) they may reman on board to perform the next flght. Otherwse, techncal and commercal crew wll board the arcraft before the start of departure ground handlng actvtes. At flght arrval, de-boardng passengers and unloadng baggage must be performed as soon as safe condtons for t are establshed so that passengers suffer as lttle delay as possble. Then accordng to the tghtness of the next departure schedule assgned to ths arcraft and the need for free parng stands, the arcraft can be drven to a remote parng poston. Unloadng/loadng of freght can be performed more or less qucly accordng to urgency and avalablty of unloadng means at the arrval parng stand or at the remote poston. Arcraft mantenance operatons, whch are n charge of the arlne and whch are not part of ground handlng may tae place, accordng to ther nature, ether at the parng stand or at a remote parng poston. Cleanng and santaton must be performed wthout too much delay to get an arcraft as clean as possble. They can be done also ether at the arrval/departng parng stand or at a remote parng poston accordng to costless and delay free opportuntes. It s also of nterest to perform potable water supply once t s possble, so that f the arcraft s requred out of schedule, only a mnmum number of ground handlng operatons wll reman to be performed. When the scheduled departure tme correspondng to the flght assgned to an arcraft approaches, the arcraft s drven f necessary to a departure parng stand. There the techncal crew (plot and co-plot) and the commercal crew get on board the arcraft. In general fuellng s realzed accordng to the arlne demand at the departure parng stand. Luggage loadng can start then untl and durng passenger boardng tme. Once fuellng, luggage loadng and passenger boardng are completed, the arcraft s ready to leave the parng stand and clearance s requested by the plot to the ATC tower. Once clearance s granted by the ATC, push bac s performed. A major characterstc of arport ground handlng s the dvers nvolvement of actvtes, from equpment, vehcles and manpower slls. Another major characterstc of 0

31 Chapter The Ground Handlng at Arport arport ground handlng s the complexty of the whole process wth parallel and sequental actvtes gong on at parng stands, transportaton lns and ground handlng vehcle bases..3. Poston of ground handlng n arport system operatons Ground handlng actvtes nteract wth arcraft traffc actvtes (taxng and apron manoeuvres) and passenger/freght handlng at termnals. Fgure.provdes a global vew of ground handlng wthn the turnaround process whle Fgure.2 llustrates n detal the poston of the ground handlng process wthn the arport system at the nterface between passenger/freght processng and arcraft arrval/departure procedures. Fgure.2 dsplays the sequencng of the man actvtes concernng wth the passenger/freght on the left, the ground handlng process as a generc module n the centre and on the rght the man actvtes concernng wth the arcraft arrvals and departures. Arrval traffc management actvtes Ground handlng actvtes Crew and Mantenance management by arlnes Departure traffc management actvtes Fgure.: Localzaton of ground handlng wthn the turnaround process

32 Chapter The Ground Handlng at Arport Unload truc Document control Baggage screenng Chec n Securty control Touch down Tax n Docng Securty control Baggage sortng Passport control GPU on Repacng Chocs on Ant-collsonlght off Ground handlng process Techncal chec FMS Wngmars Ventla -ton on Performance wx Chec rwy n use ATC clearances Load sheet Repacng Ground Handlng Load on truc Passport control Customs chec Baggage clam Ventlaton and wng mars off Ant-collsonlght on Tax out Tae off Freght Passengers and baggage Ground handlng actvtes Arcraft arrval/departure procedure Fgure.2: Arcraft related operatons at arports 2

33 Chapter The Ground Handlng at Arport Then, table.enumerates the dfferent arcraft related arport actvtes classfed nto categores dependng on where they are performed. Passenger or freght termnal Baggage chec Baggage handlng Tcetng and chec-n Passenger boardng/de-boardng Transt passenger handlng Elderly and dsabled persons Informaton systems Government controls Load control Securty Cargo Arsde Ramp servces : Supervson Marshallng Start-up Movng/towng arcraft Safety measures On-ramp arcraft servcng: Repar faults Fuellng Wheel and tre chec Ground power supply De-cng Coolng/heatng Tolet servcng Potable water supply Demneralsed water Routne mantenance Non-routne mantenance Cleanng of cocpt wndows, wng, nacelles and cabn wndows On-board servcng: Cleanng Caterng In-flght entertanment Mnor servcng of cabn fttngs Alteraton of seat confguraton External ramp equpment: Passenger steps Caterng loaders Cargo loaders Mal and loadng equpment Crew steps on all freght arcraft Table. : Scope of ground handlng operatons [Ashford and al. 203] The above representatons (Fgure., Fgure.2 and Table ;) of the ground handlng process put n evdence ts crtcal role n the turnaround process at arports and subsequently n the capacty of arports to handle flows of arcraft and passengers. 3

34 Chapter The Ground Handlng at Arport.4. Detaled analyss of the man ground handlng processes Here the most current ground handlng actvtes encountered at commercal arports are ntroduced and analysed by consderng the correspondng equpment and fleets as well as the constrants appled to them..4.. The passenger boardng/de-boardng processes At commercal arports, a boardng call on the publc announcement system ass travellers to proceed to the ext gate and board the arcraft. Boardng here s the term to descrbe the entry of passengers nto an arcraft. It starts wth allowng the entrance of passengers nto the arcraft and ends wth the concluson of the seatng of all the passengers and closure of the doors. In contrast, for the de-boardng process operatons are performed n the reverse order. Nevertheless, for both processes, arstars or arbrdges are used. Small arcraft may carry ther own stars. The boardng and de-boardng processes depend on the polcy of the arlnes (e.g. Low Cost Arlnes, Flag Carrer Arlnes) and resources avalable at a specfc arport (prncpal or remote termnals). By usng arbrdges, only the front left door of the arcraft dependng on the model s used whle by means of stars (moble stars or ntegrated stars), a second star for the rear left door of the arcraft can be used n order to speed-up the process. Hence, the operaton wth arstars s faster than the process wth arbrdges, partcularly f they are carred by the arcraft. However, ths latter statement s true only when no buses are needed to move passengers between the arcraft stand and the passenger termnal buldng. Otherwse arbrdges s more effectve and faster. These operatons are supervsed by ground personnel and cabn crew. Moreover, boardng and de-boardng can be performed smultaneously wth luggage loadng and unloadng snce these servces do not need the same area around the arcraft (n general the left sde s devoted to passengers whle the rght sde s devoted to luggage). 4

35 Chapter The Ground Handlng at Arport Fgure.3 dsplays examples of the dfferent means to board/de-board passengers. Fgure.3: Dfferent devces to handle passengers boardng and de-boardng processes.4.2. The luggage loadng/unloadng processes Checed-n luggage can be stowed n the arcraft n two dfferent ways. Ether the bags are stowed n buls or n pre-paced contaners. As the contaners can be paced before the arcraft arrves to the arport, the ground handlng process tme for loadng luggage wll be shorter wth contaner loadng than wth buls f the number of bags s large. The checed-n luggage on a flght has to be sorted, unless t s a charter flght (or other pont-to-pont flghts) where all the bags have the same prorty and destnaton. Otherwse, they mght be dvded nto transferrng bags, hgh-prortzed bags or odd sze bags and so on. Fgure.4 shows the luggage loadng/unloadng processes The cleanng process Fgure.4: Luggage loadng/unloadng processes The arlnes can request dfferent types of arcraft cleanng servces. Durng daytme the cleanng can tae from fve mnutes (tae garbage away) up to forty mnutes (garbage evacuaton, seat-pocets cleanng, belts placement, vacuum cleanng, etc.). The latter s only performed on arcraft wth longer turnaround tmes. Longer and more careful cleanng s performed durng nght-tme when the arcraft s on the ground and stay for a longer tme. 5

36 Chapter The Ground Handlng at Arport On most arcraft, cleanng and caterng can be performed at the same tme, but for some small arcraft there s not enough space for both of them at the same tme. In the latter case, t does not matter f cleanng or caterng s performed frst. The cleanng teams can proceed drectly from an arcraft to the next, but at breas and when they need addtonal materal (pllows and blanets) they have to go bac to the base. There s no sgnfcant dfference between the cleanng actvtes at dfferent arcraft types so all cleanng teams can be assgned to any arcraft type. Fgure.5 shows a cleanng team n the parng stand of an arcraft The caterng process Fgure.5: Luggage loadng/unloadng processes The caterng nvolves the wthdrawal of the leftover food and drns from the prevous flght and the supply of the arcraft wth fresh food and drns for the next flght. The caterng can start when all passengers have left the arcraft. The caterng companes use hgh-loaders to get the caterng cabnets on and off the arcraft. Hgh-loaders do not ft all arcraft types, so plannng of the assgnment of hgh-loaders to flghts s requred. The caterng process taes between fve and seventy fve mnutes dependng on how much food s needed and the way t s pacaged. The caterng teams need to go bac to the depot between servng two arcraft n order to empty garbage and get new food. The caterng coordnator maes rough estmates of the necessary manpower to perform caterng over wees and the detaled plannng, of who s servng each arcraft, are realzed every day. Fgure.6 represents two examples of the caterng process. 6

37 Chapter The Ground Handlng at Arport.4.5. The fuellng process Fgure.6: Caterng process Fuellng can be performed n two dfferent ways. At some stands there s a hydrant system wth fuel ppes n the ground that the dspenser trucs can connect to, n order to fll up the arcraft. At arcraft stands where the hydrant system s not avalable, fuellng s performed by taners. There are dfferent types of dspenser trucs: the larger types can serve all nds of arcraft whle the smaller types can only serve small arcraft. However, the small dspensers may be preferred when the area around the arcraft s tghtly lmted. Also, the tans vary n sze; n general ther capacty vares from eght to forty cubc meters of fuel. Fuellng cannot be performed smultaneously wth loadng and unloadng luggage snce these servces need the same area besde the arcraft. Before the fuel company starts to fll up, they always chec the water content n the fuel. The area around the arcraft has to be planned so that the dspenser truc or taner has a free way for evacuaton. There are also some arlnes wth specfc rules about fuellng whle passengers are on-board. Most arlnes allow t, but only under certan condtons (e.g. there must be fre extngusher ready n the mmedate surroundngs of the arcraft or there must be a two ways of communcatons between the apron and the arcraft). The tme t taes to fll up an arcraft depends on the capacty of the ppes n the arcraft and, of course, on the amount of fuel needed. The plot decdes how much fuel s needed and must report that to the fuellng company before they can start to fll up the arcraft. Today, there s no pre-planned schedule for each truc. Not untl a fuellng request arrves from the plot, the fuellng company coordnator assgns a fuellng team to t. Ths s to say that once a fuellng servce s requested, a fuellng team wll be assgned to the request and perform refuellng. Fgure.7 shows the dfferent means used to perform the fuellng process. 7

38 Chapter The Ground Handlng at Arport Fgure.7: Dfferent arcraft fuellng processes.4.6. Potable water supply and santaton process The arcraft has to be released from wasted water and re-suppled wth fresh water for the next flght. Ths s performed by two dfferent vehcles whch most often operate at the arcraft opposte sde of the luggage handlng and fuellng sde. Ths means that water and santaton can be carred out smultaneously wth luggage de-boardng/boardng and fuellng, but they must not be performed smultaneously for safety and space constrants. Fgure.8 shows the santaton process and Fgure.9 dsplays the potable water supply process. Fgure.8: Santaton process 8

39 Chapter The Ground Handlng at Arport Fgure.9: Potable water supply process.4.7. The de-cng process Snce even very thn layers of frost and ce on the arcraft have a negatve effect on the lftng force and the control of an arcraft, de-cng s needed f any part of the arcraft s covered wth snow or frost, or f there s a precptaton that could cause ths to happen. The de-cng process s dvded nto two steps: durng the frst step, frost and ce are removed from the arcraft, usually by a warm, buoyant glycol mx (type flud). The next step s called antcng and s performed to prevent new frost and ce from appearng on the arcraft before taeoff by a thcer flud (Type 2 flud). The tme from ant-cng to tae-off (called hold-over tme) s lmted, as the effect of the Type 2 flud vanshes after a whle. Ths means that t s not useful to de-ce an arcraft a long tme before tae-off. How long the hold-over tme s dependent on the type of flud, temperatures and type of precptaton. Therefore t s mportant to fnd a de-cng truc that can serve the arcraft at the rght tme. If the arcraft s served too late, the stopover tme wll ncrease wth a possble late departure as a result. If the de-cng s performed too early, the procedure mght have to be repeated. Ths result n a rather dffcult plannng problem, even f the rght tme wndows were nown n advance. Today, the de-cng coordnator plans n general on a tactcal bass consderng the current weather condtons and the flght schedule, and operatonally (when a truc s dspatched) based on a request from the plot. At the moment the coordnator gets ths request, he decdes whch truc should be assgned to the nvolved arcraft. In general, no pre-planned schedule s bult and the truc-drvers do not now n advance whch arcraft they are gong to de-ce durng the day. The request from the plot usually arrves at the begnnng of the stopover process, assumng that all actvtes wll be performed on tme. The de-cng truc wll arrve 9

40 Chapter The Ground Handlng at Arport at the arcraft some mnutes (dependng on the quantty of ce/snow/frost) before the scheduled departure tme. Fgure.0 shows how the trucs perform the de-cng operaton. Fgure.0: On-gong de-cng process.4.8. Push-bac When the turnaround process has been completed, the arcraft can depart. Arcraft at gates need to be pushed-bac usng specfc tractors. Arcraft at stands mostly requre a pushbac as well, dependng on the confguraton of the stand. At some stands, arcraft can start taxng by ts own snce the engne can be started up at the stand. The push-bac process mars a transton from ground handlng operator-arlne nteracton to ATC-arlne nteracton. Fgure. represents examples of the push- bac process. Fgure.:Push- bac process.5. Ground handlng as a complex mult-actvty process Each of the actvtes that nclude ground handlng process maes use of specalzed equpment whch must be made avalable at the arcraft parng place at the rght tme to 20

41 Chapter The Ground Handlng at Arport avod delays. Some of the ground handlng actvtes must be performed as soon as possble after the arrval of the arcraft at ther parng stand and others must be performed only at some tme before departure from ther parng stand. Dependng of arcraft operaton these two sub sets of actvtes can be performed n mmedate sequence or are separated by an dle perod of varable duraton accordng to arrval and departure schedules of a gven arcraft. Fgure.2 dsplays a standard stuaton for an arcraft undergong a turnaround process where space s a rather lmted resource and some tass cannot be performed smultaneously manly for safety reasons. It appears that the effcent operaton of such complex process whch repeats wth each arcraft arrval or departure s very dffcult to be acheved whle t s a crtcal ssue for arport operatons performance. Then advanced management tools may be useful to cope n a satsfactory way wth ths problem. Fgure.2: Arcraft servcng arrangement Typcal handlng operatons Boeng ER [Boeng, 2009].5.. Examples of ground handlng processes The ground handlng turnaround process may vary accordng to the servcng arrangement and the necessary tass for dfferent types of arcraft, dfferent operators, specfc needs for some fleets, the layout of the arport and also ts arsde management polcy. Fgure.3 dsplays the standard composton and sequencng of ground handlng actvtes for a B737. Fgure.4dsplays the composton and sequencng of ground handlng actvtes for a 2

42 Chapter The Ground Handlng at Arport medum haul arcraft at Belgrad Internatonal Arport whle Fgure.5 dsplays the composton and sequencng of ground handlng actvtes for an A320 at Stocholm Internatonal Arport. Fgure.3: Ground handlng process for a Boeng B737 [Boeng, 2009] Fgure.4: Ground handlng process at Belgrad Internatonal Arport [Vdosavljevć and al, 200] 22

43 Chapter The Ground Handlng at Arport Fgure.5: Ground handlng process at Stocholm Arport [Norn and al 2008].5.2. The temporal dmenson of Ground Handlng The turnaround (or bloc tme) s the perod of tme that the arcraft s on the arport ramp, from the blocs on at arcraft arrval to the blocs off at arcraft departure. It ncludes the postonng of the pushbac tractor and of the tow bar necessary for the push bac process. So, the turnaround perod covers all the delays necessary to perform the ground handlng actvtes as well as some dles tmes (Fgure.6). In a tght commercal operaton, mnmum turnaround wll be equal to the mnmum perod of tme necessary to complete all the ground handlng actvtes (Fgure.7) organzed n a seral/parallel process. Turnaround tme GH arrval actvtes Idle perod GH departure actvtes Fgure.6: Turnaround wth loose ground handlng actvtes 23

44 Chapter The Ground Handlng at Arport Mnmum turnaround tme GH arrval actvtes GH departure actvtes Fgure.7:Turnaround wth tght ground handlng actvtes The duraton of the turnaround wth respect to ground handlng, can tae dfferent values dependng on: The sze of the arcraft: bgger arcraft need longer turnaround tmes. For example, accordng to Arbus manuals the mnmum turnaround tme for an A320 s 23 mnutes, whle for an A340 t s 43 mnutes. It can be noted that ths mnmum turnaround tme s lower bounded by the tme requred for the braes to cool down (about 20 mnutes). The type of the flght: short-haul flghts are operated wth hgher frequency than long-haul. The short-haul flghts operate very often n tght condtons, whle long-haul flghts, whch requre longer pre-flght servcng tme, dspose n general of larger tme margns. The number of passengers or the sze of the freght to be processed. The arlne strategy: some arlnes may decde to nsert a buffer tme when plannng the turnarounds so that ther arrval/departure schedules are more robust to ground handlng unexpected delays. Arcraft bulder provde to ther customers (the arlnes) for each type of arcraft recommended ground handlng procedures tang nto account safety ssues. They produce, for each ground handlng actvty drectly related wth the arcraft, nomnal duratons as well as mnmum and maxmum values. The data stored n these charts assume standard operatonal condtons. In fact, as t was mentoned before, they are also dependent on local regulatons, on arlnes procedures and on actual arcraft condtons. 24

45 Chapter The Ground Handlng at Arport Fgure.8 dsplays nomnal duratons for the ground handlng actvtes for a B (source: Boeng 777 Manual) whle fgure.9 dsplays nomnal duratons for the ground handlng actvtes for an A (source: Arbus A330 Manual). 25

46 Chapter The Ground Handlng at Arport Passengers Servces Cargo/ Baggage/ Handlng Poston passenger Brdges or stars,0 Deplane passengers 7,5 Servce cabn - AFT LH Door 26,5 Avalable Tme Servce galleys - two trucs 29,5 Rght door no 2 Rght door no Board passengers 2,5 Remove passenger brdges,0 27,0 Rght door no 4 Unload fwd Compartment 8,0 Unload aft Compartment 4,0 unload and load bul compartment 4,0 Avalable Tme load aft compartment 4,0 load fwd compartment 8,0 Arplane Servcng Fuel arplane 23,0 Servce tolets 5,0 Servce potable water 7,0 Push bac Tme Mnutes Fgure.8: Typcal duratons of handlng operatons Boeng [Boeng, 2009] 26

47 Chapter The Ground Handlng at Arport Tme- Mnutes Deboardng/Boardng AT L Deboardng/Boardng AT L2 Headcountng Caterng AT R Caterng AT R2 Caterng AT R4 Cleanng Cargo FWD CC Cargo AFT CC Bul Refuellng Potable water servcng Tolet servcng Fgure.9: Typcal duratons of handlng operatons Arbus [Arbus, 2005] 27

48 Chapter The Ground Handlng at Arport The fgure.8shows that the total turnaround tme s about forty fve mnutes for the B and the fgure.9 ndcates that the total turnaround tme s about sxty four mnutes for the A arcraft. The fgures above are relatve to two arcraft desgned for long haul flghts. Many tass are performed smultaneously accordng to the operatons sequencngs dsplayed n the prevous secton. In the fgures, assessments are based on passengers mxed-class confguraton. It s assumed that all the equpments are worng properly and that weather condtons are normal. As the arcraft actvtes and condtons n whch these operatons are carred out are dfferent n each arport and arlne, dfferent values can be produced wth respect to the duraton of these tass Crtcal path analyss of ground handlng process It can be of nterest for managers to now for each type of arcraft nvolved n a gven ar transport operaton, what can be the best performance of ground handlng wth respect to delays. The crtcal path s the set of actvtes that are crtcal for the total duraton of the consdered process. Delayng a crtcal actvty mmedately prolongs the stopover tme. Statstcal analyses causes [Frc and al, 2009] have dentfed these crtcal processes as consstng of de-boardng, then fuellng, caterng or cleanng and fnally boardng. Accordng to the same statstcal analyses, t appears that the frequency of occurrence of fuellng on the crtcal path s 57%, 35% for caterng and 8% for cleanng. Actvtes out of the crtcal path can be delayed somehow, accordng to ther margns, wthout nfluencng the total duraton of the process. 28

49 Chapter The Ground Handlng at Arport Fgure.20: Canddate crtcal paths for ground handlng process The crtcal path of the ground handlng process vares from a flght to another snce t depends on the duraton and sequencng of the operatons. Consderng the sequencng of the ground handlng operatons on the fgure.20, a crtcal path could correspond to the followng sequences: baggage unloadng fuellng - baggage loadng or to the followng sequence: passengers de-boardng - caterng/cleanng passengers boardng or fnally to the followng sequence: santaton- potable water supply Ths wll depend on the respectve total duratons of these three paths. In the next table (Table.2), mnmal and maxmal values for the ground handlng process are produced for dfferent types of arcraft. The assumptons leadng at these values are mentoned n Annex I. 29

50 Chapter The Ground Handlng at Arport These results dsplay the large varablty of ground handlng delays n nomnal operaton..6. Concluson Arcraft Mn (mn) Max (mn) A A A A B LR B B B Table. 2: Mnmal and maxmal value for the ground handlng process The above study demonstrates the dversty and the complex nature of the ground handlng actvtes performed on a grounded arcraft whch are organzed n a seral-parallel structure where any delay on a partcular actvty may have a strong mpact on ts overall performance. Soon t appears that the dversty of actvtes to be performed as well as the need for a tght synchronzaton, not only on an arcraft but on a stream of arrvng/departng arcraft ntroduce the need for an effcent management structure to mantan ths whole process as story less as possble wthn the whole arport operatons. The effectveness of ground handlng actvtes s crtcal for arports to provde acceptable levels of servce and capacty for the processng of flows of arcraft and passengers. In the followng chapter the ssue of the organzaton of ground handlng management at arports as well as ts man objectves wll dscussed. 30

51 Chapter 2 Analyse of the Organzaton of Ground Handlng Management at Major Arport 2. CHAPTER 2 ANALYSIS OF THE ORGANIZATION OF GROUND HANDLING MANAGEMENT AT MAJOR AIRPORTS 3

52 Chapter 2 Analyse of the Organzaton of Ground Handlng Management at Major Arport 32

53 Chapter 2 Analyse of the Organzaton of Ground Handlng Management at Major Arport 2.. Introducton Accordng to the prevous chapter, t appears that ground handlng represents one of the crtcal actvtes whch s related to the qualty of servce provded by arports n handlng the flght traffc congeston there. Arport authortes, aware of ths fact, have tred n general to fnd an approprate soluton to the ground handlng management organzaton and operaton. Ths has led to a large dversty of proposed solutons wth respect to the organzaton of the ground handlng management. So, n ths chapter the staeholders nvolved wth ground handlng management at dfferent arports are dentfed, whle the pros and cons for ther nvolvement wth the ground handlng actvtes are dscussed. The relatve mportance of ground handlng wth respect to the overall management of an arport s dscussed n terms of expected costs and benefts. Fnally, the dfferent ground handlng management dutes are classfed accordng to dfferent tme scales, allowng defnng strategc, tactcal, operatonal and real tme ground handlng management functons The ground handlng staeholders When consderng dfferent arports n the world, t appears that a large varety of staeholders can be nvolved wth ground handlng management. For the dstrbuton of ground handlng functons between staeholders, there s no general standard or rule that can be appled to arports. The ground handlng operatons can be carred out under the drect or ndrect management of the followng staeholders: the arport authortes, the arlnes and specalzed ground handlng companes. Therefore ground handlng operatons can be managed globally or partally: Drectly by arport ground handlng managers, Drectly by arlnes ground handlng managers, Ground handlng companes worng for the arport Ground handlng companes worng for arlnes 33

54 Chapter 2 Analyse of the Organzaton of Ground Handlng Management at Major Arport Or by combnatons of these four stuatons. In all these stuatons specalzed subcontractors can be called to perform specfc ground handlng actvtes. The organzaton of ground handlng management at large arports depends very often on ther operatonal structure whch may nclude besdes common areas for secondary arlnes, hub termnals for man operatng arlnes. Wth respect to arport authortes, they are prmarly concerned wth the management of the nfrastructure of the arport (arsde and groundsde) to provde capacty to process arcraft traffc and passengers/freght flows. Hstorcally arports and arlnes have been nvolved n ground handlng actvtes, but wth the development of ar transportaton and the need of more and more specalzed ground handlng servces, these servces have been delegated to specalzed ground handlng companes. However, n many arports, the nvolvement of arport authortes n ground handlng actvtes remans mportant Arports, arlnes and ground handlng operators The partcpaton to ground handlng actvtes of arports authortes, arlnes and specalzed ground handlng companes present for each of them several advantages and dsadvantages whch can be determnant n many cases for the resultng ground handlng organzaton at a specfc arport. In general, the ground handlng busness s not an area from whch a consderable proft can be expected snce ground handlng staff and equpment costs are hgh whle the operaton s subject to large varatons durng a day (pea hours) and wthn the wee, wth seasonal effects whch can be very pronounced. In the case of a drect management of ground handlng actvtes by arports, revenues barely cover ground handlng costs and n many cases, they can be smaller than related costs. For the arport, these losses can be covered by revenues from other areas, such as landng fees or dverse concesson revenues. The same crcumstances happen when an arlne taes care of ts own ground handlng. Here are presented pros and cons for the nvolvement of arport authortes, arlnes and servce companes n the ground handlng sector: 34

55 Chapter 2 Analyse of the Organzaton of Ground Handlng Management at Major Arport The pont of vew of arport authortes: Advantages to partcpate n ground handlng: Master globally all the transfer processes whether for passengers/ baggage or for freght to guarantee effcent connecton and tmelness. Provde unformly to customers the requred qualty of servce by controllng and optmzng all the process flows and so mprove compettveness wth respect to concurrent arports. Ensure global safety and securty condtons by masterng smultaneously nfrastructures and processes. Provde ground handlng servces when no other staeholder s provdng t (for example the de-cng whch, beng a seasonal actvty s not attractve to nvestors). Dsadvantages to partcpate n ground handlng: Dffculty of attendng effcently the specfc ground handlng needs of the dfferent arlnes operatng at the arport, Dffculty to ntegrate and process effcently the addtonal nformaton flows generated by ths actvty. Dependng on the commercal status of some arports (publc owned), dffculty to enforce an effcent organzaton of ground handlng actvtes. The pont of vew of arlnes: Advantages to partcpate to ground handlng: Master globally the transfer processes nvolvng ther customers to ensure contnuty and tmelness of passengers, luggage or freght flows. Control the qualty of servce (delays, lost luggage occurrences, caterng, cleanness ) of ground handlng provded to ther customers to protect or mprove the arlne commercal mage. Control ground handlng operatons costs whch have an mpact on ar tcet prcng. Cover the unavalablty of local ground handlng operators or the nablty of the arport to provde t wth acceptable level of servce. 35

56 Chapter 2 Analyse of the Organzaton of Ground Handlng Management at Major Arport Dsadvantages to partcpate to ground handlng for arlnes: Ths means to localze addtonal equpment and staff at an arport whch can be a mere stopover n hs commercal networ. Ths means to be nvolved n complex logstcs problems ncludng the avalablty of ground handlng products. Penalzng constrants wth respect to the locaton and the sze of ther ground handlng depots can be mposed by the arport authortes consderng the avalable arsde areas for other staeholders. The lac of scale may turn the operaton of ground handlng by the arlne less cost attractve than when provded by a larger ground handlng operator at the arport. In some cases arlnes (arlnes allances for example) can jon together to provde a common ground handlng servce. The pont of vew of ndependent ground handlng provders Advantages to partcpate to ground handlng at a gven arport: Opportunty of proft n a large arport wth hgh levels of demand for ground handlng servces. Acqure a large share of the ground handlng maret n some mportant arports or n a networ of arports. Acqure a sound poston n arports wth hgh development perspectves n the near future. Dsadvantages to partcpate to ground handlng at a gven arport: Low proft perspectves n the near future. Strong competton of already establshed ground handlng provders. Bad operatonal condtons offered by the arport authortes. In theory, some scale advantages could be expected from centralzed ground handlng operatons. A sngle company operatng all over the arport may expect to cope wth more regular actvty levels durng the day and should mnmze duplcaton of facltes and fleets of servce vehcles. However, t can be expected that the advantages wll be balanced by the dsadvantages that come from centralzed operatons and lac of competton. Anyway the dmensons and the organzaton n dfferent areas of large arports turn n general unfeasble the dea of operatng ground equpment from a unque base. In fact, for these large arports the 36

57 Chapter 2 Analyse of the Organzaton of Ground Handlng Management at Major Arport ground handlng functon must be subdvded nto a number of self-suffcent organzatons attached to large termnals. The European Commsson has ntroduced regulatons (96/67/EC Drectve and others) to dscourage or to prevent monopoly postons for ground handlng n the European area. Here are reported the man relevant ponts of Councl Drectve 96/67/EC: Whereas ground handlng servces are essental to the proper functonng of ar transport; whereas they mae an essental contrbuton to the effcent use of ar transport nfrastructure; Whereas the openng-up of access to the ground handlng maret should help reduce the operatng costs of arlne companes and mprove the qualty of servce provded to arport users; Whereas n the lght of the prncple of subsdary t s essental that access to the ground handlng maret should tae place wthn a Communty framewor, whle allowng Member States the possblty of tang nto consderaton the specfc nature of the sector; Whereas free access to the ground handlng maret s consstent wth the effcent operaton of Communty arports; Whereas free access to the ground handlng maret must be ntroduced gradually and be adapted to the requrement of the sector; Whereas for certan categores of ground handlng servces access to the maret and self-handlng may come up aganst safety, securty, capacty, and avalable-space constrants; whereas t s therefore necessary to be able to lmt the number of authorzed supplers of such categores of ground handlng servces; whereas, n that case, the crtera for lmtaton must be relevant, objectve, transparent and nondscrmnatory; Whereas f the number of supplers of ground handlng servces s lmted effectve completon wll requre that at least one of supplers should ultmately be ndependent of both the managng body of the arport and the domnant carrer; 37

58 Chapter 2 Analyse of the Organzaton of Ground Handlng Management at Major Arport Artcle 6:. Member states shall tae the necessary measures n accordance wth the arrangements lad down n Artcle to ensure free access by supplers of ground handlng servces to the maret for the prevson of ground handlng servces to thrd partes. Member States shall have the rght to requre that suppler of ground handlng servces be establshed wthn the Communty. 2. Member States may lmt the number of supplers authorzed to provde the followng categores of ground handlng servces: - Baggage handlng - Ramp handlng - Fuel and ol handlng - Freght and mal handlng as regards the physcal handlng of freght and mal, whether ncomng, outgong or beng transferred, between the ar termnal and the arcraft They may not, however, lmt ths number to fewer than two for each category of ground handlng servces 3. Moreover, as from January 200 at least one of the authorzed supplers may not be drectly or ndrectly controlled by: - The managng body of the arport - Any arport user who has carred more than 25% of the passengers or freght recorded at the arport durng the year precedng that n whch those supplers were selected - A body controllng or controlled drectly or ndrectly the managng body or any such user The current stuaton wth respect to Ground Handlng At mportant arports such as Franfurt, Hong Kong and Genoa, the arport authorty s responsble for most of the ramp handlng actvtes as well as for passenger/baggage handlng. In that case, the arport authorty s drectly n charge of the ground handlng sector. 38

59 Chapter 2 Analyse of the Organzaton of Ground Handlng Management at Major Arport In other arports whch present major hubs for arlnes, the man ground handlng actvtes are carred out drectly or montored by these arlnes. Even, some of these arlnes can tae care of the ground handlng of other arlnes through some agreement between them. For example, USAr performs all ts ground handlng at Los Angeles Internatonal Arport and provdes ground handlng servces to Brtsh Arways. At New Yor JFK, Unted Arlnes handles not only ts own traffc but also some others from the numbers of non-u.s carrers. At some other arports, ground handlng companes have replaced arlnes to provde a servce whch was uneconomc for arlnes. For example, at Manchester Internatonal Arport, Gatwc Handlng performs all termnal and ramp handlng functons for a number of arlnes. Another example s Alled at New Yor JFK Arport, whch performs ground handlng for a number of non-based foregn carrers. Table 2. shows the results of a recent research [Norman and al. 203]concernng how ground handlng organzaton vares from an arport to another (ths research consders 72 arports from all over the world). Actvty Arport Arlnes Arport handlng company Arlne handlng company Not applcable Baggage handlng nbound 5.00% 3.00%.00% 4.00% 2.00% Baggage handlng outbound 5.69% 32.35% 0.78% 40.20% 0.98% Passenger chec-n.0% 38.53%.0% 39.53% 0.92% Transt passenger handlng 0.42% 3.25% 0.42% 34.38% 3.54% dsabled passengers servces 8.87% 30.9% 9.43% 40.57% 0.94% Ground transportaton systems 56.63% 3.6% 6.87% 2.05% 0.84% Arsde Ramp servces 26.32% 24.2% 8.42% 40.00%.05% Arsde Supervson 67.82% 0.34% 3.45% 8.39% 0.00% Arsde Marshallng 36.73% 24.49% 7.4% 30.6%.02% Arsde Start up 22.68% 28.87% 6.9% 37.% 5.5% Arsde Ramp safety control 65.96% 7.02% 0.00% 5.96%.06% Arsde On-ramp arcraft servcng 5.05% 34.4% 4.30% 39.78% 6.45% Arsde Fuellng 5.29% 4.2% 27.06% 4.8% 2.35% Arsde Wheel and tre chec 4.2% 46.39% 6.9% 4.24% 2.06% 39

60 Chapter 2 Analyse of the Organzaton of Ground Handlng Management at Major Arport Arsde Ground power supply 34.29% 22.86% 7.62% 34.29% 0.95% Arsde De-cng 3.79% 6.09% 0.34% 9.54% 0.23% Arsde Coolng/Heatng 26.60% 5.96% 8.5% 32.98% 5.96% Arsde Tolet servcng 8.56% 26.80% 7.22% 42.27% 5.5% Arsde Potable water 24.73% 22.58% 6.45% 38.7% 7.53% Arsde Demneralsed water 0.00% 7.50% 6.25% 30.00% 36.25% Arsde Exteror arcraft cleanng 6.32% 32.63% 7.37% 42.%.58% On-board servcng Cabn and 9.38% 3.25% 7.29% 5.04%.04% cocpt cleanng On-board servcng Caterng 8.05% 25.29%.49% 50.57% 4.60% On-board servcng Mnor servcng.9% 54.76% 4.76% 27.38%.90% of cabn fttngs On-board servcng External ramp 9.57% 38.30% 7.45% 38.30% 6.38% equpment provson and mannng On-board Passenger steps servcng 4.44% 30.00%.% 43.33%.% On-board Caterng loaders servcng 8.4% 26.74% 9.30% 50.00% 5.8% Table2. : Dstrbuton of responsbltes for ground handlng operatons at 72 selected arports [Ashford and al. 203] The current stuaton n Europe has been nfluenced by the 96/67/EC Drectve whose objectve was to promote for Ground Handlng effcency, qualty and prces reductons by enforcng competton between ground handlng servce provders. Ths drectve has been mplemented progressvely n the EC states and to new comng states. The man results of ths poltc have been, although arport ground handlers stll eep the majorty of maret shares, to decrease them. Also some arports have decded to sell ther ground handlng actvtes to arlnes and/or to specalzed ground handlng provders The mportance of managng ground handlng In ths paragraph, the man reasons for researchng an effcent and feasble organzaton of ground handlng at arports are revewed Ground handlng costs Ground handlng costs are supported ultmately by passengers and freght through transport fares. However arlnes have to pay for ground handlng servces whch can be seen by them 40

61 Chapter 2 Analyse of the Organzaton of Ground Handlng Management at Major Arport as fxed costs attached to a flght. In Annex II are dsplayed the ground handlng fees appled at Tallnn Arport n 202. For arlnes, turnaround costs at arports nclude all costs drectly assocated wth the servces that arlnes must pay or cover at an arport from approach, taxng, ground handlng at arrval, parng, ground handlng for departure, taxng and tae off. Then, arlnes turnaround costs nclude ar traffc control charges, landng charges, parng charges, ground handlng charges, nose and emsson charges, and passenger charges. They vary accordng to the type of arcraft and the arsde organzaton of the arport. The followng fgure (Fgure 2.) shows the turnaround charges supported at dfferent European arports (London (Heathrow Arport)- LHR, Franfurt- FRA, Venne-VIE, Munch (Frazjosef Strauss)- MUC, Madrd Barajas- MAD, Mlan Malpensa- MXP, Zurch- ZRH, Charles De Gaulle (Arport de Pars)- CDG) by an Arbus A320 arcraft Other Charges Arport charges n Euro LHR FRA VIE MUC MAD MXP ZRH CDG Securty Charges Emsson Charges Govermental Taxes Passenger Charges Parng and Brdge Charges Passenger Nose Charges Arcraft Nose Charges Landng Charges Fgure2. : Turnaround charges for an Arbus 320 at dfferent arports 203 [Zurch Arport, 203] It appears that the structure and amounts of arport charges present a large varablty n Europe. Also, snce the organzaton of ground handlng s dfferent n these arports, a varable part of these charges s destned to cover ground handlng costs. Charges drectly or ndrectly connected to ground handlng costs are: parng and brdge charges, passenger charges and securty charges, although passenger charges are manly nvolved wth passenger processng at termnals. Then, t can be consdered that n the average, no more than 5% of 4

62 Chapter 2 Analyse of the Organzaton of Ground Handlng Management at Major Arport the turnaround charges are destned to cover ground handlng costs. Ths share of turnaround charges s rather small but cannot be neglected from the pont of vew of arlnes Costs of Ground Delays for Arlnes Delay can be defned as the perod of tme to add to the scheduled tme at whch an operaton should be completed to get the actual completon tme of the operaton. Exact delay values turn avalable only once the operaton has been executed but they can be estmated n advance from dfferent probablstc models when statstcs are avalable. Of most nterest are here the delays at departure and the delay at arrval of flghts snce ground handlng can be a drect cause for departure delays, whle ground handlng may be expected to contrbute to the compensaton of delayed arrval of flghts Ground Handlng and Departure Delays for Arlnes Delay at departure can be the result of many factors and among them ground handlng malfuncton. Ground handlng delayed completon tme can result n addtonal delays when a tme wndow for tae-off, related or not wth a tme wndow for landng at arrval, s lost. Departure delays can be seen as a qualty ndex for many passengers when consderng the servce provded by the arlne and the arport. In long haul flghts, departure delays can be n many stuatons compensated by usng favourable wnds or at an addtonal fuel cost. In some other stuatons, to ths ntal delay, are added delays resultng from adverse wnd condtons. Delays at arrval result n a reschedulng of arport actvtes around the consdered arcraft. Ths s a perturbaton to any planned schedule for ground handlng whch results ether n the reschedulng of some assgnments of staff and equpment or n the actvaton of ground handlng reserve resources. There are sx man causes for flght departure delays: rotaton (late arrvals), ATFM/ATC retanng the arcraft at parng stand untl a traffc clearance s avalable, arport authortes specfc decsons (for example addtonal person/luggage checng for some securty reason), ground handlng operatons, techncal problems wth arcraft systems needng extra mantenance/repar operatons and adverse weather condtons. Observe here that rotaton delays can be caused also by upstream traffc problems coped by ATFM/ATC. 42

63 Chapter 2 Analyse of the Organzaton of Ground Handlng Management at Major Arport The table below s the results from a statstcal study of the departure delays encountered by a European domestc arlne system (Lufthansa Cty Lne) n REASON EXAMPLES PERCENTAGE Rotaton Delayed flght cycles 30% ATFM/ATC Restrctons accordng to saturated ATC sectors, traffc flow restrctons 25% Arport Authortes Problems due to lmted runway capactes, lmted 5% avalablty of parng postons, securty, etc. Ground handlng Delayed ground processes (late passengers, handlng 0% agent avalablty) Techncal problems Malfuncton of arcraft systems 3% Weather condtons Adverse weather condtons (strong ran, snow, 2% strong wnd, etc.) Other Arcraft damage, stre, communcaton problems, etc. 5% Table2. 2: Departure delay causes [Frce and al, 2009] A study performed at London Gatwc Arport n 996 (European Cvl Avaton Conference, 996) showed that the delay due to ground handlng was the second largest cause to flght delays after ATC: ATC-related delays were drectly responsble for 30% of total departure delays, whle arcraft/arlne ground servces accounted for 25% of these delays (Table 2.2). Global studes have been performed more recently n Europe and USA. The fgures bellow show results for the year 2004 where the proporton of departure delay causes are rather dfferent but demonstrate the mportance of ground delays. Ground operatons delays here nclude arlne control delays, mantenance operaton and ground handlng operatons. The dfferences n contrbuton proportons to departure delays can be explaned by the rather dfferent arspace structure and ATFM/ATC effcency, arlnes networ structure and ground operatons organzaton. Accordng to [Ronchetto, 2006], the majorty of departure delays n the US arports are the ATC n the frst place wth 37.% of the total of departure delays, the ground operatons n the second place wth 30.7% and whch nclude the ground handlng actvtes, the connecton between flghts comes n the second place wth 28.3% and the weather and the arport authortes come n the lasts places wth 3.6% and 0.2%. But t s not the case of the European arports n whch, accordng to the same study, the ground operaton comes n the 43

64 Chapter 2 Analyse of the Organzaton of Ground Handlng Management at Major Arport frst place wth 58% whch nclude the ground handlng operatons, the ATC n the second place wth 25%, the arport authortes comes n the thrd place wth %, and n the last places come the connecton between flghts and the weather wth 4% and 2% Drect cost of ground delays for arlnes The evaluaton of addtonal costs for arlnes resultng from ground delays s a dffcult ssue and dfferent fgures have been produced. When arcraft are delayed at a gate, ether wth engnes on or off, arlnes support addtonal operatonal costs and forego revenues. The overall arlnes ground delay related costs depend on the composton of ther fleet of arcraft. A study realzed by ATA for US carrers n 2004 produced the followng mean dstrbuton for departure delay causes and cost per addtonal mnute: fuel (30%, 7.05 $/mn), crew (29%, 6.77 $/mn), mantenance (8%, 0.6 $/mn, ownershp (7%, 9.74 $/mn) and others (6%, 3.36 $/mn). That means for example that 8% of departure delays was the result of late mantenance operatons wth a 0.6 $ cost per addtonal mnute. For example [Janc, 997] estmated for European arlnes the cost of a ground delay of an hour s equal to $330 for a medum arcraft, $2007 for large a arcraft and $3022 for an heavy arcraft. For the US ar transportaton maret, [Rchetta and al, 993] estmated the cost of a ground delay of an hour equal to $430 for small an arcraft, $300 for a medum arcraft and $2225 for a large arcraft. The sgnfcant varaton between these fgures can be related to the dfference of structure between the European and the US domestc networs at that tme Passengers related delay costs Delays supported by passengers represent also a cost for the arlne n two ways: - Loss of mage by offerng a perturbed transportaton servce to passengers. In general transportaton s only a mean for passengers to acheve some class of actvty (from professonal to recreatonal actvtes) and transportaton delays may have mportant consequences on these actvtes. There, complex calculatons ncludng passenger composton of flghts, wage rate dstrbuton and others, lead to dfferent fgures for the estmaton of the mean value of the lost tme per passenger and per hour. In general ths value, 44

65 Chapter 2 Analyse of the Organzaton of Ground Handlng Management at Major Arport le n other transport studes, s related wth the mean wage. For example the FAA adopted n 996 for the UK ar transportaton maret a mean value of 64 $/hour [Wu and al, 2000]. - Payment of penaltes accordng to regulatons to the passengers whch produce a clam. The delay s consdered mportant, accordng to regulaton n 26/2004 about passengers rghts of the European Parlament and Councl and assstance must be proposed to the passengers, f the flght delay s of: - two hours or more for flghts of less than 500m, - three hours or more for all (ntra-communty) domestc flghts of more than 500m and for others flghts wth dstance between 500m and 3500m, - four hours or more for other flghts. Then, when a flght has been delayed for an mportant perod of tme, the arlnes have to provde assstance n dfferent ways to the passengers: - Refreshments and possblty of restoraton dependng on the watng tme. - When the new expected departure tme s delayed for the next day, an accommodaton n hotel, the possblty to mae two phone calls/ fax and the eventual transfer to an alternatve arport have to be proposed to the passengers by the arlne. - Whatever the tnerary, f the delay s more than fve hours, the passenger are enttled to as for rembursement wthout penalty of the cost of the tcet for the part of flght not made or to flght bac to hs ntal pont of departure as soon as possble Tme Scales for Ground Handlng Management Dependng on the organzaton of arport actvtes, ground handlng management can be ntegrated to the overall management of the arport or can be performed by specfc ground handlng managers. Then, once the role of the dfferent ground handlng staeholders has been defned, dfferent tme scales can be consdered to set up ground handlng management. Fgure 2.2 presents a classcal tmelne for the management of a generc system. In the next 45

66 Chapter 2 Analyse of the Organzaton of Ground Handlng Management at Major Arport paragraph defntons for the contents of each of these management horzons n the case of ground handlng s proposed. Strategc Tactcal Operatonal Real Tme Long Term Md Term Short Term Day of Operaton Fgure2. 2: Management tmelne Strategc plannng for ground handlng The strategc plannng tme scale corresponds n general to long-term decson mang relatve to the defnton of the general phlosophy adopted for the planned system. In the case of arport ground handlng t s performed by the arport authortes and covers decsons such as the choce of ts man physcal and manageral characterstcs. For example the decson of subdvdng ground handlng by passenger termnals and some remote areas s a strategc plannng decson. The dstrbuton of ground handlng management functons between arport, arlnes and ground handler provders s another one. The structure of ground handlng charges collecton wll be also establshed at ths level (drect chargng by the ground handlng servce provders to the arlnes, ndrect chargng through arport charges, etc). Strategc plannng s based on long run predctons of traffc smlar to those used for the arport desgn plannng or upgrade. Strategc plannng provdes a worng envronment for ground handlng whch should reman roughly smlar durng some perods of operaton (several seasons or years) to provde a stable perspectve to ts ndustral staeholders Tactcal plannng for ground handlng Ground handlng tactcal plannng s concerned wth the plannng of the man resources necessary to face the demand durng the next perod of operatons for ground handlng servce. Ths s done by the managers n charge of ground handlng wthn the envronment set up by the strategc plannng decsons. At ths level ground handlng charges wll be establshed n coordnaton wth arport authortes and arlnes. Tactcal plannng s 46

67 Chapter 2 Analyse of the Organzaton of Ground Handlng Management at Major Arport performed before the start of the target perod of operaton (from three to sx months) and wth suffcent antecedence to allow the effectve avalablty of the planned ground handlng resources at the start of ths perod of operaton. These resources nclude the necessary equpment and vehcles, as well as the necessary manpower. The tactcal plannng decsons are based on medum run demand forecastng, scenaros analyss and technologcal development nformaton (new ground handlng equpment, vehcle and technques). Tactcal plannng decsons may modfy sgnfcantly the sze and composton of the ground handlng worforce through drect contractng or sub-contractng of personnel. It may nclude the tranng of personnel wth the operaton of new vehcles and procedures Operatonal plannng for ground handlng Operatonal plannng generates detaled executon plans for the next days of operaton (a wee, a fortnght). Wthn ths tme horzon, the level and composton of demand and avalable resources can be consdered nown wth suffcent relablty to start assgnng each avalable ground handlng resource to dfferent untary ground handlng demands (a flght arrval, a flght departure or both) over the perod. The problem s then to assgn the wor to each ndvdual resource as effcently as possble under the condtons specfed by the prevous plannng steps. Ths usually means, performng as many tass as possble wth the avalable personnel, whle ensurng that all operatonal constrants are satsfed. Anyway a plannng for the ground handlng operatons, amendable when necessary, s set up for the followng days Real-tme management for ground handlng Fnally, real-tme management of ground handlng operatons s concerned wth adaptng the current exstng plan for the day of operaton to handle dsturbances whch should occur durng that day. Real-tme (or dynamc) management reacts on lne to unpredcted events by reassgnng avalable resources to cover dsturbed demand for ground handlng servces. Dependng on the mportance and extent of perturbatons, ths reacton can ether be a lmted adaptaton of a nomnal operatonal plan, termed as regulaton, or a complete redefnton of t, termed as dsrupton management. 47

68 Chapter 2 Analyse of the Organzaton of Ground Handlng Management at Major Arport 2.5. Concluson The analyss performed n ths chapter shows that the concerned staeholders ( arport authortes, arlnes, specalzed ground handlng operators) are nvolved n very dfferent degrees n the management of ground handlng from an arport to another, n general accordng to specfc crcumstances. When consderng drect and ndrect costs related to ground handlng at arports, drect cost resultng from the executon of ground handlng tass represent a small amount wth respect to potental over costs resultng from even lmted turnaround dysfunctons. So, the EC recommendaton to call for ground handlng subcontractors to reduce ground handlng costs by promotng competton seems to be nessental n ths feld of actvty. What appears more mportant s the ablty of the ground handlng decson mang process to prevent dysfunctons and to reduce ther mpact when they happen. Ths ablty should operate ether at the level of the management of a specfc ground handlng actvty over an arport or at the level of the coordnaton between the dfferent ground handlng actvtes. In the followng chapter an overvew of the optmzaton approaches developed to produce effcent ground handlng decson processes at the operatons level s developed and dscussed. 48

69 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew 3. CHAPTER 3 OPTIMIZATION APPROACHES FOR GROUND HANDLING OPERATIONS: AN OVERVIEW 49

70 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew 50

71 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew 3.. Introducton The ground handlng process has receved less attenton than other arport resources management problems n the Operatons Research lterature where a rather few number of publshed wors can be found. Most of the publshed studes are focused only on one type of ground handlng resource (passenger buses, caterng vehcles, fuel trucs, etc) whle the majorty of the ground handlng management lterature copes wth off-lne stuatons. The offlne approach assumes that arcraft and arlnes meet perfectly ther scheduled arrval tmes and departure tmes, t corresponds to a stuaton where each ground handlng vehcle must be assgned to a lst of successve tass on dfferent arcraft along the operatons perod. On the contrary, n the on-lne approach a decson process must be set up to face successve or smultaneous delays on scheduled events and perturbatons n real-tme stuatons. Varants of the on-lne approach are movng tme wndow approaches and dsrupton management stuatons. Wors have been publshed wth respect to: - the management of passenger bus fleets, - the management of ol truc fleets, - the management of caterng vehcles, - the management of arcraft cleanng manpower - the management of de-cng fleets. All these problems present common characterstcs between them and wth other fleet or mult-fleet management problems found n other transportaton areas such as ndustral logstcs, dstrbuted servce delvery and port operatons. Many of these problems can be seen as off-lne arsde fleet routng problems whch may be consdered as varants of the classcal Vehcle Routng Problem (VRP) [Toth and al, 2002]. In Annex II, the man soluton approaches to the classcal VRP problem and ts varants are brefly dscussed. In ths chapter are ntroduced and analyzed some of these problems, ncludng consdered objectves and constrants, mathematcal formulaton of the problem, the proposed soluton approaches and numercal applcatons f any. Then a global analyss of the state of the art n ths feld s performed. 5

72 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew 3.2. Management of an arsde passenger bus fleet Problem defnton Here s consdered the problem of managng a fleet of arsde buses used to transport passengers from arrvng arcraft to passengers termnals and from passengers termnals to departng arcraft where n general arcraft are n remote poston and where the arcraft parng areas are lned to passengers termnals by a ground networ of lanes used n general not only by busses but also by other ground handlng vehcles. Permanent bus transportaton between passenger termnals, wth ether scheduled or unscheduled operaton wth n general larger buses, s not consdered here. The man objectve s to assgn buses to arrvng or departng arcraft so that passengers arrve on tme at destnaton (passengers termnals for destnaton passengers and departng arcraft for orgn passengers) and flghts are not delayed. Another permanent objectve s to lmt the operatons costs generated by the bus fleet by mnmzng total travelled dstances Problem class Many characterstcs of ths problem dfferentate t from other VRP (vehcle routng problems) and mae t someway harder to be tacled. Wth respect to ts specfc operatons characterstcs: - The buses operate n a pendulum way between sngle arcraft and termnals. - The followed routes are demand drven and are not repettve (no frequency of operatons). - Parng space s very lmted n the operatng area of busy arport surfaces. - The planned routes must consder possble varyng delays at the pared arcraft or passengers termnals. - The vehcles serve only one group of customers at a tme. Wth respect to the dynamc aspects of ths problem, whle t can be assumed a complete nowledge of whch arcraft (flghts) have to be servced, there s uncertanty about when and where each arcraft wll be requestng servce or how long t wll tae. 52

73 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew Problem formulaton In [Kuhn and al, 2009] the management of an arsde passenger bus fleet whch servces arcraft after ther arrval and departng arcraft before ther departure has been consdered recently. After analyzng current operatons wth the servce vehcle dspatcher at Hamburg Arport, a movng tme wndow approach was proposed by these authors where every ten mnutes an assgnment problem s solved usng updated data about the current stuaton and short term predctons. To solve successvely the resultng statc schedulng problems, a mxed nteger lnear program has been formulated n order to get current local optmal solutons mnmzng a mx of the total arcraft departure delays and of the servce provder fuel costs. The followng notatons have been adopted: Bnary varable x a j s equal to f vehcle x serves arcraft j mmedately after servng arcraft x, where =0 at the start and j=0 at the end of the servce, otherwse a j 0. Dj s the dstance a servce vehcle must travel after servcng arcraft to be ready to servce arcraft j and Dx s the dstance that vehcle x must travel from ts current poston to the poston of arcraft. T s the tme at whch the arcraft expect the servce. b s the tme at whch the servce begns on arcraft. The assumed fxed travel speed of the servce vehcle s V and Fx s the tme at whch, accordng to the current schedulng, vehcle x becomes avalable. Here I,,n and adopted: I I 0. Then, choosng a weghtng 0, x I xx I ji D a b, j, j the followng formulaton has been Mn (3.) subject to the followng constrants: x xx I a ji ji,, j I (3.2) j x a, x X (3.3) 0, j x a, x X (3.4) j, 0 x x a j a j, I I, j I,, x X (3.5) x a 0,,, j I (3.6), j b, I (3.7) T 53

74 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew D V, j I, x X (3.8) x, j x b F. a j x 0, j The frst constrant (equaton (3.2)) ensures all arcraft receve servce. Equatons (3.3) and (3.4) mpose that all servce vehcles begn and end ther servce tours at locaton 0. Equaton 3.5) s a flow conservaton constrant: a vehcle arrvng at an arcraft must leave that arcraft later. Equaton (3.7) ensures each possble tas s ether assgned or not. Equatons (3.8) provde earlest start tme constrants for the servce at an arcraft s ready Soluton approaches and comparatve results [Kuhn and al, 2009] consdered frst an exact soluton approach based on a branch and bound technque, and they compared t to a genetc algorthm, to a greedy approach and to actual operatons. These dfferent approaches were appled to problems wth 6 passenger buses servng 7 arcraft at Hamburg Arport durng an hour and to problems wth 25 vehcles servng 000 arcraft at Dallas-Fort Worth Arport durng 8 hours. At the Dallas-Fort Worth Arport, the exact soluton approach was not able to provde an optmal soluton wthn an acceptable tme. In that case, the genetc algorthm approach provded the best results over the dfferent consdered scenaros. In that case, t reduced the mean dstance travelled by the busses of about 300 lometres per day and the mean delay absorbed by arcraft by 25% relatve to the greedy approach whose performance was close to actual operatons. Then the varyng tme wndow approach, coupled wth an effcent heurstc, appeared to be able to cope rather effcently wth ths problem Management of fuellng trucs at arsde Problem defnton In many arports arcraft fuellng s performed by dedcated trucs. In large arports wth underground fuellng facltes are avalable at dec parng postons but remote parng postons must be served ndependently by fuellng trucs. In low traffc arports, n general fuellng s only performed by fuellng trucs. In general fuellng s performed only some tme before the scheduled departure tme of an arcraft. 54

75 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew Problem classfcaton Ths problem s also close to the VRP (Vehcle Routng Problem) but dffers from t by dfferent aspects: - The demand for fuel vares from one arcraft to another, mang the servcng tme dfferent. - For short turn around arcraft the tme wndow to perform fuellng may be qute reduced. - Late demands are frequent, ther orgn can be the result of new weather estmates on long haul flghts or of late adjustments n arlnes fleet operatons. - Fuellng trucs have a lmted fuel capacty and n general only one vehcle s sent to perform ths operaton at a gven arcraft. - Fuellng trucs must return to a fuel staton to recompose ther fuel load. All of ths maes ths problem to be a very specal case of VRP problem Mathematcal formulaton Ths problem has been tacled recently by [Du et al., 2008]. They studed the fuel ramp operatons and consdered the schedulng problem of fuellng vehcles and proposed a soluton approach based on the Vehcle Routng Problem wth Tght Tme Wndows (VRPTTW) wth multple objectves. Here n flghts are to be served by fuellng trucs at dfferent gates n the arport. To each flght I s attached a fuel demand d correspondng to a servce tme of duraton p and wth a tme wndow [ a, b ] wth a as earlest startng tme and b as latest startng tme. The adopted notatons are: x f truc,,m s assgned to flght y = f the th truc comes nto and and x 0 otherwse. f,,,n y =0 otherwse,,,n s s the start tme of the ground servce for flght. f,,,n. t s the flow tme of the truc, t denotes ts busy tme. 55

76 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew The objectves were n order of mportance to mnmze: m - the number of necessary vehcles: - the start tme of the servce performed by the ol tuc on each flght n order to be n able to deal wth perturbatons (accdent, flght arrval delays ): - the total busy duraton of the trucs: m y t where called nto servce, t =0 otherwse, wth C max s p. t C s B f the th truc s x and B mns. x Soluton approach Once mergng some of these objectves nto a sngle one and transformng the others n level constrants, ths problem can be formulated as a large Integer Lnear Optmzaton Problem. However t can be easly concluded that the complexty of ths resultng problem s hgh, so that heurstc approaches should be desgned to provde effcent solutons wthn an acceptable tme. Then, the authors n [Du and al, 2008] adopted a specalzed Ant Colony Optmzaton (ACO) to try to solve effcently ths mult objectve combnatoral optmzaton problem. Ant colony Optmzaton has been developed by Dorgo and al. n [Dorgo et al., 997] to solve at frst the Travellng Salesman Problem (TSP) by adoptng the collectve behavour of ant colones wth respect to food search whch s based on the current pheromone levels on the canddate trals. The heart of ths ACO algorthm s the updatng rule of the path choce probabltes. There the probablty for truc to choose flght j after havng chosen flght Is gven by: p (, j) (, j) (, j) / (, u) (, u) (3.9) u U ( ) where the postve parameters and represent the relatve mportance of the pheromone and the mpedance levels n the choce of destnaton, U () s the set of flght whch can be vsted by truc from flght I, (, j) s the level of pheromone on arc (I, j) and (, j) s the mpedance level between flghts I and j. In ths study they adopted the functon: (, j) / ( s j s ) j ( bj s ) (3.0) 56

77 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew where n the denomnator, the frst term s the travellng tme between flghts I and j, π j s the servce tme for flght j and the last term denotes the slac before the latest start tme of the servce of flght j. An Earlest Start Tme heurstc has been proposed to provde an ntal soluton, and then from one teraton to the next, local and global updatng rules have to be actvated. The local updatng rule s such that: where 0, (, j) ( ) (, j) (3.) s the pheromone decay parameter and max 0 / where 0 s computed from the ntal soluton. The global updatng rule s such that: R r max (, j) ( ) (, j) w (3.2) Where R s the set of the best solutons found at the prevous teraton and where w 0 f the r th best soluton does not use ln (,j) and w 0 otherwse. r r r r Acheved performances Numercal applcatons show that the exploraton tme of ths Ant Colony algorthm was too excessve even for medum sze problems. Then, to get better results they ntroduced an heurstc based on the Earlest Due Date. Ths heurstc h selects the flght accordng to the earlest due tme to serve when the trucs are dle. They appled ths algorthm to problems wth 20 to 54 flghts to be refuelled durng a day perod. They compared the solutons obtaned wth the above approach (lmted to 20 teratons) and an Earlest Commtted Servce Frst whch conssts n choosng the frst avalable truc each tme a flght demands refuellng. In terms of sze of the necessary truc fleet, the proposed method was best by 5% for small sze problems to 25% for larger problems, whle the computaton tmes were equvalent Management of a connectng baggage fleet Problem defnton Here s consdered the problem of managng the fleet of ground vehcles n charge of transportng baggage for connectng passengers between ther arrval and departure flghts n 57

78 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew an arport. These passengers arrve to the arport on nbound flghts and depart on outbound flghts wthn a reduced perod of tme. Ther baggages are not drected to the arrval halls le the baggage of destnaton passengers. They must be collected separately and transported to the departng flghts. The process of collectng and redstrbutng the connectng baggage vary n general accordng to many factors: the structure of the arsde ncludng termnals, parng areas and arsde crculaton lanes, the regulatons wth emphass on securty ssues and contracts between arlnes and ground operators. The handlng company s n general supposed to operate a fleet of homogeneous transportaton vehcles whch perform all the day round trps from/to the baggage dspatch faclty whle servng flghts and/or baggage handlng statons. Each vehcle returnng to the baggage dspatch faclty s assgned to a new trp whch must be performed mmedately or not, dependng of the avalablty of the baggage. Then the decson problem consdered here s relatve to the plannng of the routes for the transportaton vehcles such that each bag s delvered drectly to the flght, or to the baggage staton, respectng tme wndows constrants. The objectve s n general to delver n tme to the departng arcraft the correspondng baggage and when ths cannot be acheved wth the avalable fleet of transportaton vehcle, to mnmze the number of bags whch mss the departng flghts wthn a day perod. For example n a major European arport ths problem s handled wth two dspatch facltes whch are run ndependently on each sde of the arport (north N and south S) wth separate fleets of dentcal vehcles wth a capacty of 20 bags. Faclty N handles approxmately 4000 short transfer bags every day wth 40 vehcles whle faclty S handles about 7000 bag transfers wth 45 vehcles. There are 7 baggage handlng statons. Statstcs show that 50% of the connectng bags at faclty N are drectly delvered to the flghts whle 62% of the connectng bags are drectly delvered to the flghts at faclty S. Statstcs shown also that wth the current operaton the company has about 230 undelvered bags/day for the north faclty and about 240 undelvered bags/day for the south faclty Class of problem The baggage delvery problem s a varant of the Vehcle Routng Problem (VRP) where each delvery must satsfy strct tme wndows snce all bags for a flght must be on-board wthn a certan amount of tme before tae-off, whle they cannot be delvered untl the arcraft s 58

79 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew ready for that. Delveres to baggage handlng statons obey to maxmum delay constrants whch can be framed also as tme wndow constrants. These constrants are characterstc of a Vehcle Routng Problem wth Tme Wndows (VRPTW). However, common characterstcs to baggage delvery problems dfferentate them from a classcal Vehcle Routng Problem wth Tme Wndows: - The possblty of delverng a bag to one of two types of locatons (arcraft or baggage handlng statons) each havng dfferent tme wndow types maes ths problem be a specal Generalzed Vehcle Routng Problem (GVRP) as studed by Ghana and Improta n [Ghana and al, 2000] - The plannng of multple trps for each delvery vehcle maes ths problem be a specal Vehcle Routng Problem wth multple trps (VRPM) as studed by [Prns, 2002]. - The possblty of splttng bags between dfferent delvery vehcles for the same flght maes ths problem to be a specal case of the Mult Depot VRP (MDVRP), as studed n [Nagy and al, 2005]. Although some general framewors have been developed for large classes of Vehcle Routng Problems wth addtonal constrants [Psnger and al, 2007], [Rope and al, 2006], only the wor by Clausen and Pssnger [Clausen and al, 200] consders the whole set of the baggage delvery problem specfc constrants. In the followng, ther adopted formulaton for the offlne optmzaton problem s presented as well as the man deas of ther proposed on-lne greedy soluton algorthm wth some numercal results Mathematcal formulaton In the case consdered by Clausen and Psnger, the baggage handlng company operates a number of baggage sortng and dspatch termnals to process the connectng baggage. The company s n charge of transportng the baggage ether drectly to the departng flghts or to the baggage handlng statons where they are merged wth the other luggage assgned to the same flght. Delverng to the handlng statons s performed only f ths can be done before the bags of orgn passengers are taen from the staton to the arcraft. Ths problem has been formulated by Clausen and Psnger as a cumbersome Integer Programmng problem where N baggage must be transported usng K dentcal vehcles of capacty Q. Each vehcle s assgned to a maxmum number of routes R and each tmes return to a depot to load new baggage to be delvered wthn gven tme wndows ether at a 59

80 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew departng arcraft or at a handlng staton. A 2N+2 nodes graph s constructed where the frst N nodes represent the flghts, the next N nodes represent handlng statons, node 0 represents the ntal depot whle node 2N+2 s the fnal depot. Let V be the set of nodes and E be the set of edges. Strong connectvty s assumed for ths graph. An empty route connects drectly the ntal and the fnal depots. To each node s assgned a tme wndow [a, b ]. The processng duraton at node I s gven by s whle the travel tme between two nodes I and j n the graph s gven by t j and the arrval tme of baggage s wrtten u. The bnary varable x f vehcle goes from to j along the r th route, x 0 otherwse, z f baggage s not delvered and z 0 f t s delvered, Sr s the tme at whch servce at node s completed by vehcle on route r. Then we get: N Mn z (3.3) Subject to jr x x z,, j V, K, r R (3.4), j,, r n, j,, r x, j V, K, r R (3.5) 0, j,, r x, V, K, r R (3.6), 2n,, r x j,, r Q,, j,, j 2N, K, r R (3.7) x x 0, V, K, r R (3.8), j,, r j,,, r jr x j r Sr s tj S, j,,, jr,, K, r R (3.9) a S b, V, K r, r R (3.20) S S, K 2nr 0r (3.2) x j,,,, r S0r u, V, K, r R, j V (3.22) x 0,,, j, K, r R (3.23), j,,, r 60

81 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew z 0,, (3.24) S 0, V,, r, K, r R (3.25) Relaton (3.3) consder the objectve of mnmzng the number of undelvered bags whle constrant (3.4) sets z to f bag s not delvered on tme to ts flght or handlng staton, Constrants (3.5) and (3.6) are depot startng and endng condtons (each route should leave the depot once and return to t once). Constrant (3.7) s a vehcle capacty constrant whch must be satsfed on all routes. Constrant (3.8) s a flow conservaton constrant at node for vehcle performng route r. Constrant (3.9) ensures that f edge (, j) s used by vehcle on route r, then the completon tme at j s greater than the departure tme at node plus the travel tme between and j and drop off tme at j. Constrants (3.20) are tme wndows constrants and constrants (3.2) nsure that new routes cannot be started before the prevous routes have ended. Constrant (3.22) ensures that vehcle cannot start route r untl ts correspondng baggage s avalable Proposed soluton approach Consderng the sze of real lfe nstances and the dynamc aspect of the problem, a greedy algorthm was proposed to solve approxmately ths problem. Wth ths algorthm, each vehcle s scheduled ndvdually and only for one trp at a tme. The schedulng s performed once a vehcle arrves to the dspatch hall (at start of ts operaton or when t returns from a prevous delvery trp). Then at that tme a delvery tas s generated and assgned to the drver of that vehcle. Ths tas ndcates whch set of bags must be pced up at each locaton n the dspatch hall and the lst of delvery destnatons for each bag. The algorthm s desgned so that good sets of tass are generated. A good set of tass has been defned as beng such as flghts wth an mmnent departure flghts are treated wth prorty, the tas assgned to a vehcle should handle as many bags as possble and the routes assocated wth the delvery tass should be as short as possble. Then the proposed heurstc maes use of penalzatons to handle these sub-objectves. The algorthm consders all bags present n the dspatch hall at the tme of calculaton and the nduced sub-graph contanng only nodes and Edges belongng to the depot. For each edge 6

82 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew (, j) s computed a cost that should reflects the attractveness of delverng the bag assocated wth node j: cj (3.26) where s the cost assocated wth the type of delvery, L j L L j R R j D D j R j s the cost assocated wth the D length of the route and j s the cost assocated wth the departure tme at locaton j, here,, are real valued weghts. L R D The edges wth lowest cost are selected n a greedy way up to delvery tme constrants or vehcle capacty constrants Obtaned results To test the algorthm, they used real data about transfer bags for a full wee of operatons. The arport consdered n ther tests was composed of two dspatch facltes. The numercal results showed that the proposed algorthm s robust wth regards to the stochastc aspect of the bag delvery tmes and the vehcle travel tmes Management of a de-cng fleet Problem descrpton Arcraft de-cng becomes a necessary ground operaton before arcraft departure when there s t has been pared for some tme n cng condtons and there s a rs that a layer of ce forms on the arcraft crtcal surfaces. In that case the arcraft aerodynamc effcency can be largely deterorated and a tae-off manoeuvre wthout de-cng can lead to a crash stuaton. The de-cng operaton s consdered to be curatve when ce has been already formed and the assocated ant-cng operaton s consdered to be preventve snce the effect of the ant-cng lqud remans for a tme suffcent to tax and tae-off safely. De-cng s n general the last ground operaton before taxng for tae-off. The de-cng process can be centralzed at de-cng statons or decentralzed wth the use of a de-cng fleet of vehcles. The need for de-cng s dependent on actual weather condtons and arcraft state. Conservatve decsons are n general taen by consderng meteorologcal 62

83 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew forecast, but current condtons can turn ths operaton unnecessary for some flghts. That means that the demand for de-cng cannot be establshed too much n advance snce t presents can present a large degree of uncertanty. It s worth to observe that the duraton of the de-cng operaton wll depend on the mportance of the arcraft cng state Current studes [Norn et al, 2009] developed a smulaton model for the assessment of the turn-around actvtes of a de-cng fleet at an arport. Ths model was valdated usng Stocholm Arport as reference arport. Then they proposed a mathematcal formulaton of the de-cng fleet schedulng problem where the objectve s to mnmze a mx of the total delay for the departng flghts and of the total dstance travelled by the de-cng vehcles. Ths modellng approach s detaled n the next paragraph. [Mao and al, 2008], consdered the case of an arport wth de-cng statons to whch arcraft have to go to be processed before departure. They vewed ths problem as a specal case of a Mult-mode Resource-Constraned Project Schedulng Problem (MRCPSP) [Bruer, 999] where the objectve s to mnmze the total delay of arcraft at tae-off. There the arcraft were taen as agents and the de-cng statons as resources. A pure Frst Come Frst Served-FCFS heurstc has been compared wth a FCFS heurstc ncludng penaltes (decommtment penaltes-dc) to promote the coordnaton between agents and mae them reserve the de-cng trucs as close as possble to ther tae-off tme. The results show that comparng the FCFS to the FCFS wth DC, the second approach gves a lower delay regardless of the number of arcraft. [Zhwe and al, 200] proposed another Mult-Agent based model for the schedulng of arcraft de-cng operatons. They try to show that the mult-agent approach [Feber, 995] can be useful n managng ths problem by allowng to tae better nto account the uncertanty and flexblty of the problem and to preserve the nterest of all the concerned actors (the arport, the arlnes and the ground servce company). They proposed a decson mang algorthm based on the negotaton between agents whch proved superor to a mere FCFS strategy n terms number of de-ced arcraft per perod and n the arcraft de-cng delays. 63

84 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew Mathematcal formulaton Here we consder the mathematcal formulaton proposed by [Norn et al, 2009] for the decng fleet schedulng problem. It s as follows: subject to N Mn N 0 N K R j0,j r r N r x h xhj 0, h 0,, N N K j R j0 r r a. l b. w. x (3.27),,,K, r,, R j j (3.28) x, 0,, N (3.29) r j N N r j d. xj q,,,k, r,, R (3.30) j r xj t j t s f w M,, j,, N,,,K, r,, R (3.3) p t s f,,, N (3.32) p,,, N STD (3.33) l t s f STD,,, N (3.34) t stop m f start z t 0 M mn n, m, n,, R,,,K (3.35) m n z x x, n m,, j 0,, N,,,K (3.36) mn 0 0 j stop r tr p j wj0 M x j0, j,, N,,, K, r,, R 0 start r tr t w0 M x0 (3.37),,, N,,, K, r,, R t 0, p 0, 0 l,,, N (3.38) (3.39) Here K s the number of avalable de-cng trucs; N s the number of assgnments durng the consdered tme perod. M s an arbtrary large constant. Assgnment 0 s to the truc fuel staton where also all routes start and end; R s the total number of routes performed by the trucs. A route s a feasble sequence of assgnments for a fuel truc. R s chosen large 64

85 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew enough to accommodate all the routes that the fleet can perform n a day ( R N ), note that some of these routes may be empty. Here q s the capacty of truc, w s the travellng j tme for fuel trucs between assgnments and j, f s the mean de-cng duraton tme, f s 0 the truc refll tme at the fuel staton, STD s the scheduled departure tme of arcraft and s start s the de-cng set up tme at arcraft, t s the start tme of fuellng at assgnment, t s the stoprt start tme for the route r, t s the stop tme for the route r, p s the end of the tme r assgnment, l s the delay for the arcraft correspondng to the assgnment. r a and b are the weghts of the objectve functon (total servce delay at arcraft and total truc, travellng tme, respectvely). r Wth respect to decson varables, the adopted notatons were such as: x f there s an arc r from to j on route r for the truc, otherwse x 0 ; z f the truc performs the route m before the route n, otherwse z 0. mn Then, equaton (3.27) s the objectve functon whch corresponds to the mnmzaton of a weghted mx of the delay of arcraft resultng from the fuellng servce and of the total travellng tme of the fuel trucs. Equaton (3.28) ensures that the same trucs arrves to and leaves each assgnment on ts route. Equaton (3.29) defnes that every assgnment s performed exactly once. Equaton (3.30) maes sure that a de-cng truc s gong to the refll staton before t runs out of flud. Equaton (3.3) specfes that a truc cannot arrve to an assgnment before the prevous one s completed and the truc has travelled between the assgnments. The tme an assgnment s fnshed s calculated n equaton (3.32) and (3.33). The possble flght delay s defned n equaton (3.34). Equaton (3.35) defnes that the next route wth the same truc cannot start before t s re-equpped wth de-cng flud. Equaton (3.36) guarantees that f an arc exsts (.e. f the x-value for an arc s ) the z-value for the correspondng route s also. Equaton (3.37) and (3.38) specfes the start and stop tmes for a route. j mn j 65

86 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew Soluton approach and results The above problem s a mxed lnear optmzaton problem (bnary varables r xj and z mn, real varables (t, p ) whose soluton, even for small sze nstances, requres a large computatonal effort. Then, to get worng solutons to ths problem n an acceptable computaton tme, dfferent GRASP (Greedy Randomzed Adaptve Search Procedure) heurstcs [Feo and al, 995] were developed. These technques generate durng the search process a set of concurrent solutons from whch domnatng solutons wth respect to the two man objectves are retaned. A smulaton model was used to compare n the case of Stocholm Arport the de-cng operatons performances resultng from a GRASP based management and from current schedulng rules. They used data from Stocholm Arport before and after the ntegraton of the concept of Collaboratve Decson Mang and n both cases the GRASP approach proved superor to current schedulng rules Management of caterng fleets Problem descrpton A more sophstcated soluton was proposed by [Ho and al, 200] to tacle the arlne caterng operatons ncludng the staff worload. They consdered the problem as a manpower allocaton problem wth tme wndow and job-sll constrants. The optmzaton objectve conssts n the maxmzaton of the total number of assgned jobs. They presented a comparson between Tabu Search and Smulated Annealng approaches to solve the problem. To test these approaches, they used real-lfe nstance provded form an arlne caterng company. The results show that the Tabu Search gves better solutons than the Smulated Annealng approaches. They studed also the mpact of the team formaton and they found that the extenson of allowng jobs to be shared between two teams s a good mode of operatons Mathematcal formulaton [Ho and al, 2009] consdered a flght as a job. There are n jobs by the set J,,n, where each job s descrbed by an arcraft/ confguraton combnaton 66 f F, a servce

87 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew duraton p and a, whch denotes the earlest and latest startng tmes for a job. b, s a set of d drvers, and L,,l D,d s a set of l loaders n a day, where the total worers s the setw D L. Worer W s descrbed by hs/her shft hourst,, and a set of lls represented by arcraft type/confguraton combnatons, and worer j and worer j j W S F e. Worer, may have overlappng slls,.e., S S. All worers must travel n teams when leavng the depot, denoted by 0, when vstng job locaton, the team returns to the depot, denoted by n (although physcally located the same as 0 ). Teams are formed by groupng drver and loader j together, where t t and e j j j e, D, j L. It s assumed that the number of loaders n a shft s at most the number of drvers n the same shft. Loader j must be n a team wth a drver, whereas, drver mght be n a team wth drver h, where t t and e h h e,, h D. Hence, there are m teams, denoted by,, where nsh Dq Lq m L q q 2 V,m day. L denotes the set of loaders n shft q, q be served by team 67. Here, nsh denotes the number of shfts n a D denotes the set of dvers n shft q. Job v can q V (wth members and j ) f f S S (.e. at least one of the two j team members has the requred sll), s a, b and s t and s p e, where s denotes the start of servce for job. The overall manpower schedulng problem conssts of constructng a set of team, teams-to-jobs assgnment and job start-tmes such that a balanced schedule whch mnmzes the number of unassgned jobs s made.m and M 2 are arbtrares large constant matrces. The set L j L t t, e e same shft as drver D L j D t t, e e each worer varable x s defned as: j / for D, s defned as the set of loaders who are n the j j j D. The followng sets are defned n smlar manner: / for L j / for D. For D j and D j D j, t t, e e D D and worer j L D, and for each team V, the decson, f worer and worer j belong to team j (3.40) 0, otherwse x For each par of job locatons u and v, whereu v J 0, n, u v the decson varable y uv s defned as: j,, and for each team, j

88 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew, f team does job v "mmedately"after job u y (3.4) uv 0, otherwse Note that job 0 refers to the ntal departure from the depot and job n refers to the fnal arrval at the depot. To model the job-slls compatblty constrants, an ndcator parameter each job s defned for v, f job v s n sll - set of worer v J and each worer W as: (3.42) v 0, otherwse The decson varable s s defned for each job u u and denotes the start of servce of job u (by some team ). The basc manpower schedulng problem can be started mathematcally as: Max (3.43) V u 0 J vj y uv Subject to: V L D x, j L (3.44) j j V jl D x j x, D (3.45) D x xj 2 xj, V (3.46) D j L jd D D x x, j D, V (3.47) j j y, V (3.48) 0v vj n u u y y, V, h J (3.49) uh hv 0 J vj n 0 y, V (3.50) u, n J V u 0 y, v J (3.5) J uv 68

89 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew u 0 J v vj y uv x, V, v J (3.52) D, jw j a v s b, v J (3.53) v v t x j yuvm D s, V, u 0 J, v J, D, j L D (3.54) v s v p v yuvm e xj M 2 D D V, u J 0, v J, D, j L (3.55) s u yuvm p s, V, u, v J (3.56) u v s 0, u J (3.57) u D x 0,, V, D, j L D (3.58) j y 0,, V, u, v J 0, n (3.59) uv Constrant (3.44) restrcts the team assgnment wth a loader to drver of the same shft, whle constrant (3.45) states that a drver mght be grouped wth ether a loader or a drver of the same shft. Constrant (3.46) ensures that no more than two worers are assgned to each team (ndex). Constrants (3.48)-(3.50) guarantee that for each trp the team leaves the depot, after servcng job n sequence, t fnally returns to the depot. Constrant (3.5) states that each job s assgned to at most one team. Constrant (3.52) states that job v could only be served by team f job v s ether n the sll-set of worer or n the sll-set of worer j. Tme wndows constrants for job v are specfed by (3.53). Inequaltes (3.54) and (3.55) specfy that f team s vstng job locaton v, ts servce duraton must fall wthn the shft hours of team. Constrant (3.56) ensures that servce perods between trps of team are ordered sequentally. (3.58) and (3.59) are the nternalty constrants. The objectve (3.43) s to maxmze the number of assgned jobs (n realty, t s also mportant a balanced schedule, and t has been addressed n the soluton methodology). 69

90 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew 3.7. Global approaches Recently, some authors have consdered the global arport ground handlng schedulng and assgnment problem. The global approach has been tacled n two man ways: a fully centralzed approach and a fully decentralzed approach. The wor by [Dohn and al, 2008] has concentrated on the management of ground handlng manpower by consderng that ground handlng s managed by a central entty responsble to buld up dynamcally the teams wth the dfferent nvolved slls, whch wll be n charge of each arrvng or departng arcraft. The decentralzed soluton approach of the global ground handlng assgnment problem has been coped n two ways: - by consderng that the global ground handlng schedulng problem s an nstance of a mult-project schedulng problem, - by consderng that t s a dstrbuted decson mang problem A Centralzed Approach for the Ground Handlng Assgnment Problem Problem descrpton Here t s consdered that each ground handlng demand (arrval, departure or both) s processed by unts composed of equpment/vehcle and specalzed manpower. Servce delvery at arrvng or departng arcraft obeys to tme constrants whch can be expressed as tme wndow constrants. Then when followng a partcular ground handlng team, t s successvely assgned to dfferent servces at dfferent locatons and performs a tour whch covers some of the parng stands wth grounded arcraft. Then t can be consdered that each ground handlng unt performs a sub tour whle t s expected that the whole grounded arcraft wll be vsted by the requred teams of ground handlng operators. 70

91 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew Mathematcal formulaton [Dohn and al, 2008] proposed a formulaton of the schedulng problem of personnel at arports where the objectve s to mnmze the total number of unassgned tass and mnmze the operatng cost of each team. So they ntroduced the Manpower Allocaton Problem wth Tme Wndows whose formulaton s as follows: Let C,,n be a set of n tass and consder a set V of nhomogeneous teams of worers. To each tas s assocated a duraton, a tme wndow, a set of slls and a locaton. It s supposed that each tas C has to be performed n a tme wndow a, b where a and b correspond to the earlest and the latest startng tmes for a tas. Each tas s dvded nto r splt tass. Tme t j s the transportaton tme between each par of tass, j and the servce tme at tas. If team has the requred qualfcatons for performng tas, then g otherwse g 0. Each team V unque servce centre at locaton 0, common to all teams. operates wthn a worng tme wndow e, f from a The selected objectve s here to mnmze the total number of unassgned tass whle assgnng to each team feasble sequences of actvtes along paths. Such feasble paths are shfts startng and endng at locaton 0 and obeyng at tme wndows and sll requrements constrants. They are defned by the sequence of tass they vst. Let x f tas j s performed drectly after tas by the team and x 0 otherwse. s s an nteger varable and defnes the start tme of the cleanng on the arcraft. Subject to: x j V C jn j Max (3.60) j xj V jn r, C (3.6) x, C, j C, V (3.62) j g jn N x 0 j, V (3.63) x h xhj 0, h N, V (3.64) jn 7

92 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew e toj M 0 x j s j, j C, V (3.65) s x f t M 0 0, C, V (3.66) j xj s j s t M, C, j C, V (3.67) a s b, C (3.68) x 0,, N, j N, V (3.69) j 0 s, C (3.70) The objectve (3.60) s to maxmze the number of assgned tass. A tas s counted multple tmes f t s processed by more than one team ( r 2 ). The constrants (3.6) guarantee that to each tas s assgned at most the rght number of teams or possbly less, f some of ts splt tass are left unassgned. Only teams wth the requred sll can be assgned to a specfc tas (3.62). Furthermore, constrant (3.63) s used to ensure that all shfts start n the servce center. Constrants (3.64) ensure that no shfts are segmented. Any tas vsted by a team must be left agan. The next four constrants deal wth the tme wndows. Frst, a team can only be assgned to a tas durng ther worng hours (3.65) (3.66). Next, the tme needed for travellng between tass s avalable (3.67). If a customer s not vsted, the scalar M, whch has been chosen arbtrarly large, maes the correspondng constrants non-bndng. Constrants (3.68) enforce the tas tme wndows. Fnally, constrants (3.69) (3.70) are the ntegralty constrants. The ntroducton of a servce start tme removes the need for sub-tour elmnaton constrants, snce each customer can only be servced once durng the schedulng horzon because t j s postve. The formulated problem s NP-Hard Soluton approach [Dohn and al, 2008] consdered that ths problem s close to the vehcle routng problem wth tme wndows. So they adopted a Column Generaton technque assocated wth a Branch and Bound technque, resultng n a Branch and Prcng approach [Desaulners and al, 2005]. Here the soluton approach s based on the consderaton of feasble paths, where a feasble path s a shft startng and endng at the manpower base. An nteger master problem has been ntroduced to assgn to each team a feasble path so that the total number of assgned tass s 72

93 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew maxmzed, but the synchronzaton between the tass cannot be drectly tacled. The selected objectve s here to mnmze the total number of unassgned tass whle assgnng to each team feasble sequences of actvtes along paths. Such feasble paths are shfts startng and endng at locaton 0 and obeyng at tme wndows and sll requrements constrants. They are defned by the sequence of tass they vst. When an optmal soluton s not obtaned (soluton s not nteger or tas synchronzaton constrants are not met) a branchng s performed accordng to the soluton of a prcng problem. Here the prcng problem results n elementary shortest path problem wth tme wndows for each team whch are solved usng a label settng algorthm Applcaton to the management of cleanng manpower Arcraft cleanng s essental n order to mantan the hgh qualty standards of servce delvered on-board arcraft by the arlnes to the passengers. Dependng of the way the arcraft s operated (long haul flghts, fast connectons for domestc/regonal arcraft) the requred servce can ether be tghtly constraned by tme slots or not and these tme constrants can ether be nown wth a large antcpaton or not. In general cleanng (and tolet refurbshng) s performed once arrvng passengers have left the arcraft and before departng passengers arrve. In general at the gate the ground personnel of the arlne chec that cleanng s completed before allowng passengers to board the arcraft. Dependng on the parng poston of arcraft (at gate or remote) ground vehcles are necessary to transport the cleanng teams to the arcraft. [Dohn and al, 2008] llustrated ther approach to optmze manpower allocaton for ground handlng wth the case of the arcraft cleanng manpower at an arport. To evaluate for that applcaton the of effectveness ths approach, test data sets taen from real-lfe stuatons faced by arlne cleanng companes n two European major arports have been used. The test data set has been organzed n four dfferent problem types and each type has been composed of three problem nstances coverng 24-hour perods. From 0 teams and 00 tass up to 20 teams wth 300 tass have been consdered. The authors reported that the above exact soluton approach has provded effectve results for the smallest nstances after computaton tmes spannng from seconds to hours whle tme out or memory out stuatons have been obtaned wth larger nstances. Then ths exact soluton approach, whch leads to numercal dffcultes 73

94 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew n the off-lne stuaton, wll not be of nterest n the on-lne context unless heurstc procedures are ntroduced to replace ts exact search processes Decentralzed Approaches of the Global Ground Handlng Assgnment Problem Mult-project schedulng approach A representatve wor for ths approach s the one of [Mao and al, 2009] whch proposed a soluton to solve the arport ground handlng schedulng problem under uncertanty by consderng that the global ground handlng schedulng problem s an nstance of a multproject schedulng problem (MPSP), so, they consdered the arcraft as a project agent whch s composed by a set of actvtes, and the ground handlng provders as resource agents, each one s responsble of a resource whch performed a specfc type of actvty. As a frst step, they provded a formal descrpton of ths nstance tang nto account the uncertanty at the level of the executon tme of the operatons. The second step, and n order to cope wth the uncertanty, they proposed an onlne mult-agent schedulng approach. In ths approach, they presented an onlne schedule based on a cooperatve scheme. It has been noted that ths approach could only handles the uncertanty at the level of the release tme and t was dffcult to apply t n the case of the presence of dsrupton n the processng duraton. That why, n the thrd step, n order to deal wth the dfferent nd of dsruptons, they proposed to use the same structure (MPSP) to nsert slac tme between the actvtes. Ths slac tme would guarantee, n case of the appearance of any ncdent that the resources stll wor as planned. The frst approach was appled n a determnstc envronment, usng 0 type of arcraft turnaround procedures, for each procedure there were 0 dentcal arcraft nstance. The results obtaned by the applcaton of the two mult-agent schedulng approaches: noncooperatve and cooperatve, were been compared wth 3 centralzed heurstcs methods: Frst Come, Frst Served (FCFS), Maxmum Total Travel Wor Content Frst and Shortest Actvty from the Shortest Project. The results showed that for the fve schedulng approaches the total project delay (turnaround tme) decreases wth the ncrease of the delay cost per tme unt. From computng tme pont of vew, the Maxmum Total Travel Wor Content Frst and the Shortest Actvty from Shortest Project heurstcs methods had the shortest computng tme. Concernng the resource levellng measures, t has been observed that the mult-agent 74

95 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew schedulng wth the cooperatve scheme used to carry out the ground handlng processes the lowest resource levellng. So, accordng to the results, the cooperatve onlne schedulng scheme was the one of the best centralzed schedulng heurstcs. For the second proposed approach, n order to calculate the adequate slac tme to nsert n the end of each actvty, a genetc learnng algorthm was employed. Ths approach was appled for dynamcs problems (resources neffcency). The results showed ths approach was able to absorb the delays at the level of the executng tme of actvtes, to converge to a stable stuaton and to avod reschedulng the resources Dstrbuted decson mang approach Followng ths approach, [Garca et al, 20] consdered the ground handlng processes as a dstrbuted decson support system. To deal wth ths problem, they created a new theoretcal and expermental Mult-Agent System called MAS-DUO. The archtecture of ths new MAS was based on a combnatons of many exstng methodologes. The MAS-DUO s a dvson of the organzaton model n two platforms: system of nformaton model and physcal model. Each platform was treated ndependently to better understand the system and to facltate the desgn and the development of the MAS. Ths dvson allowed strategc polces to be reflected on the physcal decsons and nformed to the upper nformaton system about physcal dstrbuton as well. The communcaton between the two platforms was assured by usng of an nteracton protocol based on sharng parameters of the Marov reward functon. Ths new organsaton was tested to manage the ground handlng operatons on the Cudad Real Central Arport. The ground handlng operatons tang nto account corresponded to the set of operaton performed on a Boeng B737 durng a standard 45 mnutes scale Analyss and concluson The consdered applcatons of Operatonal Research to solve ground handlng operatons problems at the operatons level, treat n general a nomnal problem wth no perturbaton to the arcraft arrval schedule or to the operatons of the dfferent ground fleets. Even n ths nomnal case, the correspondng mathematcal programmng problems are of hard complexty class wth bg dffcultes to get exact solutons for real sze problems. Then, some heurstcs have been bult to provde a soluton to these nomnal problems. In general heurstcs of the greedy type can be adopted to cope wth on lne perturbatons snce they treat 75

96 Chapter 3 Optmzaton Approaches for Ground Handlng Operaton: An Overvew n sequence the dfferent decsons to be taen. However few wors report some experments where the heurstc appled to ground handlng schedulng are assessed n perturbed envronments. Wth respect to the mult-agent approaches, they focuses manly on the mnmzaton of the costs supported by each ground handlng agent whch are consdered at the same level than delays supported by passengers. In the frst class of studes an actvty-based decentralzed organzaton of ground handlng s adopted mplctly but no coordnaton scheme s proposed. In the second class of studes, the ntensty of nformaton flows necessary to process maret-based mechansms or perform mult-agent based decson mang s such that a centralzed approach appears preferable. Then t appears that the majorty of these studes mssed two cornerstones of the consdered global ground handlng operatons problem: The cost dmenson, whch has been consdered n the prevous chapter and where t s clear that the drect cost resultng from ground handlng actvtes are secondary wth respect to the economc consequences of delays at servcng arrvng and departng arcraft. The management dmenson where an organzaton able to cope wth routne stuatons as well as perturbed condtons or even dsrupted stuatons, must be desgned. In reference to ths last pont, n the followng chapter, the desgn of an effcent organzaton of ground handlng management compatble wth global approaches to cope wth nomnal, perturbed and dsrupted stuatons at arports s developed. 76

97 Chapter 4 A Global Organzaton of Ground Handlng Management 4. CHAPTER 4 A GLOBAL ORGANIZATION OF GROUND HANDLING MANAGEMENT 77

98 Chapter 4 A Global Organzaton of Ground Handlng Management 78

99 Chapter 4 A Global Organzaton of Ground Handlng Management 4.. Introducton In ths chapter the problem of the organzaton of ground handlng management wthn an Arport Collaboratve Decson Mang (A-CDM) envronment s explored. Frst the man A-CDM prncples are recalled and the level of nteracton of ground handlng nformaton wth the whole arport management through A-CDM s dscussed. Snce ground handlng actvtes generate very large flows of dfferentated nformaton and accordng to the A-CDM mlestone approach, a two level structure for the management of ground handlng, where the upper level nteracts drectly wth the other A-CDM partners, s nvestgated. Then the functons to be developed by a ground handlng coordnator (GHC) at the frst level and the specalzed ground handlng managers (GHMs) at the second level are dscussed. Petr nets are ntroduced to represent and analyze the logcal structure of these functons as well as the coordnaton processes adopted between them A-CDM and ground handlng management The A-CDM concept The objectve of the concept of A-CDM, ntated by the European Commsson n 2008, s to enhance the overall effcency of the European Ar Transport System. Ths overall effcency s consdered achevable f the ar and the ground segments of ths system operate n harmony. Then, accordng to traffc estmates provded by the ar traffc servces (ATFM, ATM, ATC), arports operatons should present a hgh degree of predctablty. Ths s acheved by performng arport actvtes wthn accurate tme tables. The arport partners nvolved n the A-CDM are then: Ar Traffc Control (ATC), arcraft operators (manly arlnes), ground handlng management, ar traffc networ management and arport operatons managers. Fgure 4. dsplays all the A-CDM partners and the nteracton between them. 79

100 Chapter 4 A Global Organzaton of Ground Handlng Management AIR Traffc Control (ATC) Traffc Networ Management Arport slots &pre-departure sequence Fght data processng Fght plan/ other data Common data bases Fght plan/ other data Arcraft Operators Arport Operators Informaton systems Ground Handlng Fgure 4. :The arport partners nvolved n the A-CDM The concept of A-CDM s manly based on the followng general prncples: - Share at the rght tme of relevant data between the dfferent partners. - The qualty of the exchanged data must contrbute to the predctablty of events and the plannng capablty of decson maers. - Interface decsons are assgned to one of the nvolved partners. - All partners are nformed on-lne of the adopted decsons. The applcaton of these prncples should mprove the effectveness of decsons of each decson maer, where objectves and constrants of other decson maers are consdered together wth ther actual and predcted stuatons. These prncples are the base of the man functons of the A-CDM whch could be summarzed n these four ponts: - Mlestone approach - Arcraft process executon assessment - Trend analyss of the pre-departure sequence - Arcraft process status The Arport CDM s supported by an nformaton sharng system composed of computer networs, databases and user nterfaces. The structure and scope of ths nformaton sharng system depend on the organzaton of the arport and ts staeholders. 80

101 Chapter 4 A Global Organzaton of Ground Handlng Management Operatonal prncples of CDM The operaton of A-CDM s based on two man operatonal prncples: - The collaboratve management of flght updates: the flght arrval nformaton s provded by the ar traffc networ management to the CDM arport whch provdes smultaneously flght departure nformaton to the ar traffc networ management. The coordnaton between Ar Traffc Flow and Capacty Management and arport operatons of a CDM Arport should mprove the effcency of the ATFM slot management process for departng flghts. - The adopton of a mlestone approach whch descrbes the progress of a flght from the ntal plannng to the tae-off by defnng sgnfcant events to be closely montored. Bloc-off and tae-off are among the most sgnfcant events. The adopton of ths approach should enhance the tme predctablty of the followng events for each flght. To produce accurate and effectve predctons about departng traffc, arport ATC should provde arcraft ground traffc nformaton to all CDM partners: - Frst, tax-n and tax-out delays are computed (varable tax tme) to mprove the estmaton of the boc-n and tae-off tmes, ncreasng then the ground traffc predctablty. - Second a pre-departure sequencng provdng the order n whch arcraft are planned to depart from ther stands (bloc-off, push bac) s communcated to the other partners. Ths sequence must ntegrate constrants and objectves of the other partners to nsure feasblty and mprove slot adherence. The adopton of these operatonal prncples should enable the arport to cope as effcently as possble ether n normal stuatons (good weather condtons, no capacty lmtaton) or n adverse condtons Ground handlng and A-CDM As a result of the mproved predctablty of arcraft arrval tmes at parng stands, ground handlng management can expect to acheve: - An enhanced punctualty of ground handlng operatons. - The agreement wth requred ground handlng servce levels. 8

102 Chapter 4 A Global Organzaton of Ground Handlng Management - The mnmzaton of ground handlng operatons costs. The mproved predctablty should allow the ground handlng managers to antcpate the necessary resources needed by an arrvng arcraft and moblze at the rght tme the rght ground handlng resources. Here, bloc-n nformaton wll be provded on the medum range by the ar traffc networ management and on the short run by the arport ATC tower, whle the arcraft operator wll nform about the specfc ground handlng servces requred by the arrvng or departng arcraft. However, the ground handlng process presents some mportant specfc characterstcs wthn the arport operaton: - It s a process nvolvng dfferent resources (equpment and manpower) managed n general separately. - The ground handlng process may vary n composton accordng to the characterstcs of ts operaton. - The duraton of the dfferent ground handlng tass may vary even for the same type of arcraft accordng to ts occupancy. Then, the ground handlng process s a potental generator of an enormous flow of nformaton of whch only a small part s relevant to the global objectve of mprovng traffc fludty and safety wthn the ar transportaton system. It does not appear convenent to communcate all ths nformaton to all arport partners (too much nformaton lls nformaton). In the next paragraph, accordng to an adopted overall organzaton of arport ground handlng, mlestones wll be proposed for the followng up of ths actvty Introducng an Arport Ground Handlng Coordnator When consderng ground handlng organzaton n dfferent arports, t appears that ths organzaton depends strongly on the sze and the physcal organzaton of the arsde as well as on the volume and composton of traffc. Then, as shown n chapter 2, a large dversty of actual ground handlng organzatons s found n major and medum sze arports. Then t does not appear desrable to propose a general paradgm to organze arport ground handlng snce the resultng effcency can be qute unequal from an arport to the next. 82

103 Chapter 4 A Global Organzaton of Ground Handlng Management However, when some ey characterstcs are met, delmtng a specfc class of ground handlng stuatons, common organzng prncples can be of nterest. Here some assumptons wth respect to arport ground handlng characterstcs, whch are frequently encountered n medum to large arports, are adopted. They are the followng: - Here s consdered the case of arports n whch ground handlng s performed by a set of specalzed operators worng n parallel under the management of the arport authortes. - The ground handlng process s supposed to follow pre-establshed sequencngs and to be performed at the parng stands. - It s supposed that the parng stands are assgned to arrvng flghts by the arport and communcated through ATC, whle the status of the parng stands s montored by ATC whch s n charge of drvng the arcraft out of the parng poston. - It s also supposed that the arrvng parng poston s ts departure parng poston for the next flght. Ths last assumpton ntroduces constrants on the ground handlng actvtes. From the consderatons developed n the prevous paragraph, t appears nterestng to consder that the arport ground handlng operators do not nteract drectly wthn the A-CDM framewor, but through a ground handlng coordnator (see fgure 4.2). Ths coordnator wll nterface the other arport partners wth the ground handlng operators: - The coordnator wll provde each ground handlng operator of ground traffc predctons and requred ground handlng resources for each flght. - The coordnator wll provde the other arport partners wth predctons of ground handlng delays and mlestones completon nformaton. 83

104 Chapter 4 A Global Organzaton of Ground Handlng Management A-CDM Real arrval tmes Arlnes ATC Scheduled arrval and departure tmes GH Coordnator Scheduled start tmes Real completon tmes Ground handlng Manger Real start tmes Scheduled completon tmes Ground handlng Manger j Ground Handlng Unt Document Ground Handlng Untj Arcraft Arcraft +p Fgure 4.2 : Connecton of A-CDM wth Ground Handlng In ths stuaton, the GHC should drectly exchange data wth the followng A-CDM partners: - ATC/ATM: to get predcted tmes of arrval of arcraft at parng poston. It s supposed that the choce of the parng poston has been solved and nformed through a drect exchange between ATC/ATM and the correspondng arlne. - Arlnes: to get nformaton about the effectve ground handlng needs of arrvng/departng arcraft. The GHC wll be able to provde to the arlne a predcton of completon tme of ground handlng actvtes at arcraft arrval/departure. Then the arlne wll be able to communcate wth ATC/ATM and negotate departure tme f necessary Ground handlng mlestones montorng by GHC The ground handlng actvtes around an arcraft can be dvded n two set of operaton: - The set of arrval ground handlng operatons, gh A, whch ncludes all the ground handlng actvtes whch must be performed to conclude properly the current 84

105 Chapter 4 A Global Organzaton of Ground Handlng Management commercal flght. The man arrval ground handlng actvtes are de-boardng passengers, unloadng baggage, performng cleanng and santaton. gh - The set of departure ground handlng operatons, D, whch gathers the ground handlng actvtes whch must be performed to prepare the next commercal flght. The man departure actvtes are passengers boardng, baggage loadng, fuellng, caterng. Ground handlng at arrval Tax n Unload luggage De-boardng passengers Santaton Cleanng Fuellng Caterng Load luggage Boardng passengers Potable water supply Ground handlng at departure Pushng bac Tax out Fgure 4.3 : Example of the set of ground handlng actvtes for an A320 at Stocholm arport Fgure 4.3 represents an example of ground handlng actvtes sequence for an A320 at Stocholm arport and how those actvtes are dvded n two sets. To lmt the flow of nformaton sent to the other A-CDM partners, t appears that the nformaton about the startng and end tmes (planned and effectve) for arrval and departure 85

106 Chapter 4 A Global Organzaton of Ground Handlng Management ground handlng actvtes s suffcent to manage predctablty of operatons at the overall arport level. Then the possble mlestones montored by the ground handlng coordnator are for an arrvng flght operated by arcraft : agh - tme of start of arrval ground handlng actvtes T whch s such as: T agh gh A agh mn t (4.) agh - tme of completon of arrval ground handlng actvtes whch s such as: agh gh A agh agh max t d (4.2) Here agh agh t s the start tme of ground handlng actvty on arrvng arcraft, d s the duraton of ground handlng actvty on arcraft. In the same way, the possble mlestones montored by the ground handlng coordnator are for a departng flght operated by arcraft : dgh - tme of start of departure ground handlng actvtes T whch s such as: T dgh gh D dgh mn t (4.3) dgh - tme of completon of departure ground handlng actvtes whch s such as: dgh gh D dgh dgh max t d (4.4) Here dgh dgh t s the start tme of ground handlng actvty on departng arcraft, d s the duraton of the ground handlng actvty on arcraft. All these tme related varables and parameter adopt two values: ther estmated value whch can evolve and ther effectve value at completon Ground Handlng Coordnaton In ths approach, besdes montorng mlestones for the beneft of the other A-CDM partners, the Ground Handlng Coordnator (GHC) coordnates the dfferent ground handlng fleets whch operate smultaneously at dfferent places of the arport. 86

107 Chapter 4 A Global Organzaton of Ground Handlng Management Ths central manager receves through the A-CDM updated nformaton about predcted flght arrvals and flght departures and dstrbutes ths nformaton to the dfferent ground handlng managers. These specalzed ground handlng managers provde hm n return wth effectve start and completon tmes, so that he can produce completon mlestones nformaton (on-lne estmatons and fnally effectve values) to the A-CDM partners. Observe here that the A-CDM approach can be of nterest to organze the flows of nformaton between the specalzed ground handlng managers and the ground handlng coordnator, but also between them. Ths wll lead to the concept of GH-CDM as a sub nformaton networ dedcated to mprove ground handlng effcency (Fgure 4.4). Fgure 4.4:Introducng a Ground Handlng CDM In chapter 2 t was demonstrated that effcency n ground handlng actvtes s characterzed manly by the tmelness of the process (arrval or departure ground handlng, arrval and departure for short turnovers), whle the costs resultng from ground handlng nvestment (fxed and moble equpment) and operatons costs (staff, fuel) present a much lower mportance. 87

108 Chapter 4 A Global Organzaton of Ground Handlng Management To acheve tmelness n an envronment such as the arsde of an arport characterzed by mportant uncertantes nherent to ar transportaton, the ground handlng process should be able n some crcumstances to speed up, perform the whole arrvng and/or departure ground handlng n mnmum tme, accordng to some crtcal path technque [Clare and al, 2004], and then recover some of the ntal delay. Crtcal path technques assume mplctly that the necessary resources to perform the dfferent actvtes (ether on the crtcal path or not) are on the spot ready to be used. Then, the search for an effcent ground handlng supposes the avalablty of the correspondng resources (equpment and staff). Here t s proposed that the ground handlng coordnator s n charge of the global plannng of ground handlng resources whle ground handlng operatons are performed n a decentralzed way by each specalzed ground handlng manager or GHFM (ground handlng fleet manager) accordng to ths resource requrement by the GHC. It appears of nterest to perform globally the estmaton of ground handlng resources snce n ths way, synchronzaton between dfferent ground handlng actvtes s drectly taen nto account n the computaton and the adopted resources margns follows a sngle approach. The presence of these planned margns for the ground handlng resources wll prevent from delay propagaton over long perods of tme. These ground handlng resources should be computed once the schedule of arrvals and departures s avalable for the next day. Also, when a major dsrupton occurs at the arport wth needs for fast recovery towards regular operaton, temporary capacty problems may appear as the result of an unexpected out of proportons ncreased level of demand, ncludng for ground handlng processng. In that case t s expected that the ground handlng coordnator wll tae over ground handlng actvtes by enforcng prortes decded at the A-CDM level Global plannng of ground handlng resources The plannng of ground handlng resources should be performed at start for a whole day of operaton by consderng as basc nput nformaton: - the tme schedule of arrvng and departure flght, - the operatonal characterstcs of these flghts. 88

109 Chapter 4 A Global Organzaton of Ground Handlng Management Ths nformaton wll be provded respectvely by the arport authortes and the arlnes. Also a pre-assgnment of arcraft to the dfferent parng areas of the arport s supposed to be avalable. Ths pre-assgnment can be produced perodcally by the arport authortes n agreement wth the nvolved arlnes. When large ar traffc perturbatons happen, the ground handlng coordnator wll decde to update the plannng of ground handlng resources by consderng the predcted demand for ground handlng servces durng a shorter perod of tme. Ths shorter plannng perod wll be taen long enough to allow the return to nomnal condtons. Ths approach can be extended to the management of major dsruptons by tang nto account explctly, as ntal constrants, the current ground handlng stuaton. The soluton of the global ground handlng plannng problem wll allow hm to perform a predcton of the necessary amount of ground handlng resources (vehcles and wor force) need at each tme perod. Ths predcton wll be acheved n three steps: - At the frst step, a global ground handlng assgnment (GGHA) problem s solved for a nomnal schedule of flghts. Here the objectve s to mnmze the sum of the delays for the completon mlestones of the ground handlng of each flght. Ths problem wll be consdered n detal n the next chapter and a fast heurstc soluton wll be proposed. Ths soluton wll produce wth respect to each ground handlng operator a set of nomnal feasble routes from one arcraft to the next so that each foreseen ground handlng tas wll be covered by a vehcle from the correspondng fleet at the rght tme. Ths nformaton can be forwarded to some ground handlng fleet n some crcumstances, but n general t wll have a lac of robustness wth respect to perturbatons and may soon turn unfeasble. - At the second step, totalzaton of necessary resources s performed. It s consdered that the whole operatng perod s composed of dscrete tme perods. A unt tme perod equal to the maxmum between 5 mn and the smallest duraton of a ground handlng operaton, ncludng travel tmes between parng stands and depot, can be adopted. Then consderng the feasble routes produced by the soluton of the GGHA problem durng a gven perod for a specfc ground handlng fleet, summng provdes the nomnal estmaton of the necessary resources of ths type durng that perod of tme. 89

110 Chapter 4 A Global Organzaton of Ground Handlng Management - At the thrd step, margns are added to the estmaton of necessary resources. Here, to the prevous estmaton, margns are added to mprove the avalablty of ground handlng resources n front of perturbatons. There s no exact method to compute these margns to provde some probablty of success snce the dstrbuton and composton of perturbatons s not n general characterzed n probablstc grounds. However some basc prncples can be consdered. Need for extra resources are the result of unexpected peas of demand. Snce n general no antcpaton s allowed n normal operatons condtons, ths pea of demand for ground handlng servces at a gven perod can only be created by the accumulaton of delays (ether arrval or departure delays) n the near precedent tme perods. Based on avalable delay statstcs for arrvals and departures the formulaton of a stochastc global ground handlng assgnment problem, where the objectve would be to mnmze the mean value of total delays resultng from ground handlng whle lmtng the sze of the nvolved ground handlng teams at each tme perod, wll be extremely complex. A possble determnstc way could be to modfy the nomnal schedule before a gven tme t by ntroducng delays just before ths tme, for example a 20 mnutes delay at arrval or departure for arcraft scheduled to arrve or to depart wthn the prevous half an hour. Then the global assgnment problem wll be solved wth ths modfed schedule leadng to an estmaton of necessary resources at tme t. Ths process should be repeated all over the dfferent tme perods composng a day (24 2=288 tmes). Ths approach s too cumbersome, even f, as t wll be decded n chapter V, the global ground handlng assgnment problem wll be solved usng a greedy heurstc. Then a smpler approach than the above approach can be to consder at a gven tme the resources necessary to meet the nomnal arrval and departure schedule and, consderng the nomnal traffc durng the prevous half, add accordngly some margn. A smple rule could be such as: For arrval ground handlng actvtes: r n p A (4.5) A 90

111 Chapter 4 A Global Organzaton of Ground Handlng Management where: - n s the nomnal number of teams (vehcle and staff) of type necessary at perod to process scheduled arrvals. - r s the computed requred number of teams of type necessary at perod, to process schedules arrvals, ncluded reserve, - A s the number of teams of type necessary to handle flght arrvals at parng stands durng the prevous half an hour whch are supposed to be processed before perod and - p A s the probablty that an arrval scheduled wthn half an hour before perod s delayed and should be processed at perod. For departure ground handlng actvtes: r n p D (4.6) D where: - n s the nomnal number of teams (vehcle and staff) of type necessary at perod to process departures. - r s the computed requred number of teams of type necessary at perod to process departures, ncluded reserve. - D s the number of teams of type necessary to handle flght departures at parng stands durng the prevous half an hour whch are supposed to be processed before perod and - pd s the probablty that a departure scheduled wthn half an hour before perod s delayed and should be processed at perod. For arrval and departure ground handlng actvtes: r n p A p D (4.7) A D 9

112 Chapter 4 A Global Organzaton of Ground Handlng Management where: - n s the nomnal number of teams (vehcle and staff) of type necessary at perod to process arrval and departures. - r s the computed requred number of teams of type necessary at perod to process arrvals and departures, ncluded reserve. Observe that the computaton of these ground handlng resources does not nclude the spare vehcle stoc whch should be dmensoned, accordng to statstcs, by the ground handlng manager, to guarantee a gven relablty level. The ground handlng coordnator wll choose the values of probabltes p A and accordng to other factors such as weather and season. p D accordng to the avalablty level he targets and A - A n - n Fgure 4.5 : Plannng of a safe level for ground handlng resources In the Fgure 4.5, a smple descrpton of how the plannng of a safe level for ground handlng resources for each ground handlng manager s presented Decentralzed ground handlng management Decentralzed ground handlng management wors at two complementary levels: the local level and the coordnaton level Local ground handlng management Each ground handlng manager GHM, = to T, where T s the total number of ground handlng actvtes, has to manage fleets of vehcle and people to mae them avalable once they are necessary to perform the ground handlng actvtes they are n charge. Then to mae that possble at lower costs (nvestment, operatonal costs), each ground handlng manager has to manage dfferent bacground actvtes. Some of these bacground actvtes are plannng actvtes performed on the long-medum run and related wth fleet and 92

113 Chapter 4 A Global Organzaton of Ground Handlng Management manpower dmensonng and acquston/recrutng. Other bacground actvtes, performed on the medum run, nsure fleet mantenance and acquston of supples necessary for the ground handlng actvty (chemcals, water, ndustral food, etc) as well as vehcle fuellng. Consderng that ar transportaton at an arport s not present durng a wee an overall perodcty, to acheve ts ground handlng msson, a GHM has to solve on a daly bass the assgnment of hs resources to the ground handlng tass whch are affected n a temporal bass to hm by the ground handlng coordnator. Instead of solvng an ntegral assgnment of manpower ndvduals and specfc vehcles, ths problem s splt nto two assgnment subproblems. At the upper level a parng problem s consdered by the GHM where the objectve s to assgn the avalable ground handlng unts ( GHU ) to ground handlng tass of type wth the objectve to mnmze ground handlng servce delays whle mnmzng drect operatons costs. These ground handlng unts or teams, are n general composed of an equpped and suppled vehcle and a team of operators. These drect operatons costs are related to the ntensty of use of ground handlng unts and to the total dstance travelled by the correspondng ground handlng vehcle. Ths problem wll be referred as the ground handlng fleet assgnment (GHFA, = to T) n the next chapter. At the lower level, ground handlng unts are bult up from the stoc of worng vehcles and avalable manpower. A ground handlng unt can be n the followng states: - deactvated: ether the equpment s not ready (under repar or mantenance) or the operators are not avalable, - watng for assgnment: the unt s enabled but has not been assgned to flghts, - assgned: the unt has been assgned to one or more flghts, but the realzaton of the actvty on the frst of these flghts s planned far n the tme horzon, - made ready to perform ts next actvty: ths happens when the planned tme to perform a ground handlng actvty s near. Ths corresponds ether to the tme necessary to adapt the resource to the flght to be served or to a mnmum tme delay to nform the operators of the next operaton, - operatng: the unt s performng the actvty (transfer operatons and processng at arcraft or termnal). 93

114 Chapter 4 A Global Organzaton of Ground Handlng Management Wth respect to manpower, once the parng problem has been solved, ndvdual assgnment can be performed n two steps: - The frst step s performed on a tme bass, where to each partcular employee s assgned, or not, an actvty perod. Durng ths perod the employee s ether worng effectvely wthn a ground handlng team at ground handlng tass or he s ready to start a new tas. Then, personalzed ground handlng teams are bult up. - In the second step, these personalzed ground handlng teams are assgned to the ground handlng tass through the soluton of the parng problem. At both steps, regulatons wth respect to worng condtons must be met. One of the man objectves of these regulatons s to enforce safe worng condtons to avod accdents. In ths thess only the parng problem wll be consdered explctly snce from the effcency of ts soluton wll depend drectly the performance of the arport whle the consttuton of the ground handlng unts should reman transparent to the A-CDM partners Coordnaton level of ground handlng management To be at least feasble, a decentralzed approach, nomnal or on-lne, must be coordnated n some way snce each ground handlng tass must be solved accordng to a sequence compatble wth the need of ground handlng actvtes for a partcular arrvng or departng arcraft. In the nomnal case where arcraft arrve at and leave from the parng stands on schedule, stuaton whch happens scarcely, the planned sequence of actvtes at the parng stand could be adopted to solve successvely and n parallel the dfferent GHFA problems, the solutons of the upstream GHFA problems provdng earlest startng tme constrants for the downstream GHFA problems. However, any perturbaton wll mpar the effcency of the whole ground handlng performance. In general arcraft at arrval use to be ether n advance, on tme or delayed dependng on traffc and wnd condtons. Here, to cover all these stuatons, t wll be supposed that ground handlng resources assgned to an arrvng arcraft should be ready to start operaton from ther respectve base wth some antecedence wth respect to scheduled arrval tme at the gate. Dependng f the flght s a short, medum or long haul, ths antcpaton wll be smaller or larger. In the case of departng arcraft n commercal operaton, n general there wll be no 94

115 Chapter 4 A Global Organzaton of Ground Handlng Management antcpated departure, so the effectve departure schedule wth eventually some delay, wll be the bass for ground handlng operatons at arcraft departure. The central manager whch receves through the A-CDM updated nformaton about predcted flght arrvals and flght departures wll be able to provde on-lne to the dfferent ground handlng managers the start tme nformaton assocated wth each upcomng flght Petr Net representaton of proposed ground handlng organzaton and operaton One am of ths part s to develop a model of the proposed ground handlng organzaton n order to nvestgate ts senstvty to the occurrence of dfferent types of dsruptons as: changes of avalable resources (arcraft stands- gates, equpment, personnel, etc.), arcraft arrval delays, as well as dfferent gate assgnment strateges. Consderng the concurrence, precedence constrants and synchronzaton aspects of ground handlng actvtes, Petr nets appear to be of nterest to model ths stuaton snce Petr Nets are nown to be a powerful tool to model and smulate dscrete systems nvolvng all the aspects of the ground handlng process. Also, snce tme plays an mportant role n the performance of ground handlng systems, Tmed Petr Nets appear of specal nterest here. The ground handlng organzaton can be modeled by consderng the three operaton and management levels as shown n fgure 4.6: - Ground handlng unts 2- Ground handlng manager 3- Ground handlng coordnator GHC GHM GHM j GHU GHU j GHU j GHU j Fgure 4.6 : Three-levels organzaton of ground handlng management 95

116 Chapter 4 A Global Organzaton of Ground Handlng Management Ground handlng unts The Ground handlng unts belongng to a specalzed ground handlng provder have to communcate wth other agents whch are a part of the ground handlng system: - It has to be able to communcate wth ts ground handlng manager to provde hm the state of the processng of the tas (start tme, completon tme, on tme, occurrence of any dsrupton, equpment falure). - It wll also receve from ts ground handlng manager new assgnments at other parng postons or passenger or luggage statons n the arport. - It has to alert the watng ground handlng unts of the completon of ts tas at the arcraft. The followng RdP ( Fgure 4.7) represents the dfferent operatonal states of the GHUs wth the nformaton whch s exchanged durng the processng of ther ground handlng tas. Fgure 4.7 : RdP representaton of GHU s operatons 96

117 Chapter 4 A Global Organzaton of Ground Handlng Management Here the nterpreted places and transtons are as follows : - P0: the GHM assgns the ground handlng unt j (GHU j ) to perform the ground handlng tas at locaton, or tas -. - P: the GHU j s assgned to perform a ground handlng tas - and s ready to start t. - P6 up: s a data sent by the GHU j, whch performs the upstream ground handlng tass at the same staton, to the GHU j representng the followng state: the upstream ground handlng tass to - have been already completed on tme, accordng to the scheduled completon tme, by the GHUs n charge of them. - P9 up: s a data sent by the GHU j, whch performs the upstream ground handlng tass at the same staton, to the GHU j representng the followng state: the upstream ground handlng tass to - have been already completed on tme wth a delay accordng to the scheduled completon, by the GHUs n charge of them. - P2: the GHU j starts to perform the ground handlng tas -. - P3 : s a data sent by the GHU j to hs GHM representng the followng state: the GHU j starts to perform the ground handlng tas - wth a delay accordng to the scheduled start tme. - P4 : s a data sent by the GHU j to hs GHM representng the followng state: the GHU j starts to perform the ground handlng tas - on tme accordng to the scheduled start tme. - P5: an ncdent has happened durng the executon of the ground handlng tas -; t results n a delay for ts completon tme. - P6 : s a data sent by the GHU j to hs GHM representng the followng state: the GHU j has completed on tme the ground handlng tas - (accordng to the scheduled completon tme). - P9 : s a data sent by the GHU j to hs GHM representng the followng state: the GHU j has completed wth a delay the ground handlng tas -. - P7: the GHU j equpment s n a faled state and the GHU j operators are unable to complete the ground handlng tas -, t has to be replaced by another one. 97

118 Chapter 4 A Global Organzaton of Ground Handlng Management - P8: a new equpment s ready to replace the faled one and to perform untl completon the ground handlng tas -. - P0: the GHU j has already fnshed performng the ground handlng tas - and he s avalable to be assgned to perform another ground handlng tas. - P7 : s a data sent by the GHU j to hs GHM representng the state P7. - P6 down: s a data sent by the GHU j to the GHUjwhch perform the downstream ground handlng tass at the same staton representng the followng state: the GHU j has completed on tme the ground handlng tas - (accordng to the scheduled completon tme). - P9 down: s a data sent by the GHU j to the GHU j whch perform the downstream ground handlng tass at the same staton representng the followng state: the GH Uj has completed wth delay the ground handlng tas -. - T0: ths transton allows the GHU j to pass from the state avalable to assgned to perform the ground handlng tas - due to the decson made by the GHM. - T: ths transton allows the GHU j to start performngthe ground handlng tas - snce they are ready and the upstream ground handlng tass are completed wth a delay accordng to the scheduled start tme. - T2: ths transton allows the GHU j to start performng the ground handlng tas - snce they are ready and the upstream ground handlng tass are completed on tme delay accordng to the scheduled start tme. - T3: t s a tmed transton, the tme represent the end of the tas. The completon tme of the ground handlng tas, n ths case, s represented by an nterval n whch t was consdered the earlest completon tme and the latest completon tme. - T4 : f the GHU j has not fnshed the ground handlng tas yet ( T3 has not been fred), n ths case, the GHU j s not on tme, and a delay appears at the level of ths tas - T5 : t represents the end of performng the ground handlng tas after the occurrence of the delay 98

119 Chapter 4 A Global Organzaton of Ground Handlng Management - T6: t represents the event that the delay s caused by the falure of the GHU j equpment. - T7: the falure GHU j equpment has been replaced by the reserved one and the GHU j can contnue to perform the ground handlng tas. The places P0, P3, P4, P6, P9, P8 and P7 represent a communcaton nterface between the GHU j and hs GHM. The places P6 up and P9 up represent a communcaton nterface between the GHU j and the upstream GHUjwhch perform the upstream ground handlng tass at the same staton. The places P6 down and P9 down represent a communcaton nterface between the downstream GHUjwhch perform the downstream ground handlng tass at the same staton and the GHU j. The places P, P2, P5, P7 and P0 are the dfferent states of GHU j durng the processng of the ground handlng tas, that s why they have been sent to the GHM to have an overvew of what happens for each GHU j Ground handlng manager The ground handlng manager must have a detaled vew of what happens at the level of each of hs ground handlng unts. Also, he has to communcate data to the ground handlng coordnator. The followng RdP (Fgure 4.8) represents the dfferent operatonal states of the GHUs wth the nformaton flow sent to the GHM durng the processng of ther ground handlng tas and how the GHM uses t to assgn each GHU to each ground handlng tas. It represents also how the GHM ntervenes n case of a GHU s equpment falure. 99

120 Chapter 4 A Global Organzaton of Ground Handlng Management Fgure 4.8 : RdP representaton of operatons by a GHMj Here the nterpreted places and transtons are as follows : the places P0, P3, P4, P6, P9, P8 and P7 represent, as mentonned before, a communcaton nterface between the GHU j and hs GHM. P3, P4, P6, P7 and P9 are the mage of what happens realy durng the processng of the ground handlng tass. The GHM taes nto account these states to assgn each GHU to each ground handlng tas. T7: f a GHU equpment s n falure and the GHM has spare equpment, n ths case ths transton can be fred Ground handlng coordnator The ground handlng coordnator must have a global and detaled vew of what happens at the level of each of hs ground handlng manager. Also, he has to communcate data to the A- CDM. The followng RdP (Fgure 4.9) represents the communcaton between the GHC and the GHMs on one sde and the other partners of A-CDM on the other sde. 00

121 Chapter 4 A Global Organzaton of Ground Handlng Management Fgure 4.9 : RdP representaton of operatons by a GHC Each Ground Handlng Manager (GHM ) has to send the real start and completon tmes of hs ground handlng actvtes performed on each flght to the Ground Handlng Coordnator (GHC). After recevng these data the GHC can start to calculate the mlestones of the arrval and departure actvtes and send them to the A-CDM. The data sent to the GHC by the GHM: - P3 : s a data sent by the GHU j to hs GHM representng the followng state: the GHU j starts to perform the ground handlng tas - wth a delay accordng to the scheduled start tme. - P4 : s a data sent by the GHU j to hs GHM representng the followng state: the GHU j starts to perform the ground handlng tas - on tme accordng to the scheduled start tme. 0

122 Chapter 4 A Global Organzaton of Ground Handlng Management - P6 : s a data sent by the GHU j to hs GHM representng the followng state: the GHU j has completed on tme the ground handlng tas - (accordng to the scheduled completon tme). - P9 : s a data sent by the GHU j to hs GHM representng the followng state: the GHU j has completed wth a delay the ground handlng tas - (accordng to the scheduled completon tme) Concluson In ths chapter t has been shown that adoptng a herarchcal approach, t s possble to organze ground handlng management n accordance wth the A-CDM approach where a ground handlng coordnator operates as an actve nterface between the ar transportaton operators (arport authortes, ATC and arlnes) and the specalzed ground handlng managers n charge of the ground handlng unts. In ths organzaton the ground handlng coordnator generates to the other A-CDM partners the mlestones assocated wth ground handlng and provde to each ground handlng managers safe values for the ground handlng resources necessary to face not only nomnal stuatons as well as perturbed ones. Accordng to ths approach, ether the ground handlng coordnator as each specalzed ground handlng manager faces decson problems. In the next chapter, the soluton of these decson problems wll be consdered through the consderaton of the correspondng optmzaton problems. 02

123 Chapter 5 Decson Mang Processes for the Proposed Global Approach 5. CHAPTER 5 DECISION MAKING PROCESSES FOR THE PROPOSED GLOBAL APPROACH 03

124 Chapter 5 Decson Mang Processes for the Proposed Global Approach 04

125 Chapter 5 Decson Mang Processes for the Proposed Global Approach 5.. Introducton In ths chapter the man decson mang processes n charge of the manageral unts composng the proposed ground handlng management organzaton n the prevous chapter, are consdered. The adopted approach s here to formulate a correspondng optmzaton problem, to propose eventually an exact soluton approach and chec ts practcal feasblty and then to propose a possble heurstc approach. The objectves adopted for these optmzaton problems concentrate on the respect of global or local tme deadlnes wth some consderaton for the correspondng operatng costs, accordng to the analyss performed n chapter II. The constrants ntroduced n the respectve formulatons nsure that the resultng solutons are physcally feasble when consderng the nvolved dscrete resources and the spatal and temporal dmensons of these problems. The generated optmzaton problems are at least partally combnatory, ths mples n general long processng tmes. Then, the heurstcs approaches are of partcular nterest snce t s of utmost mportance to be able to get practcally onlne updated feasble solutons when perturbatons occur. In ths study t s consdered that every tme an arcraft operatng a flght drected towards a gven arport taes-off, that arport s nformed of ts departure as well as ts predcted landng tme. The predcted landng tmes can be updated durng the flght The Central Planner Problem The frst decson problem consdered here s relatve to the szng of resources performed by the ground handlng coordnator (GHC) to be sure that durng daly operatons, the dfferent ground handlng managers (GHMs) wll have the necessary resources n equpment, vehcle and people to cope wth nomnal operatons as well as perturbed stuatons. Ths problem, whch tacles globally the dfferent ground handlng actvtes, s supposed to be solved ndependently by the GHC. Consderng the dffcultes ponted out n the prevous chapter to solve n some optmal way ths problem whch has also some stochastc characterstcs, t has been proposed to solve t n two steps: whle n a frst step an 05

126 Chapter 5 Decson Mang Processes for the Proposed Global Approach overall nomnal assgnment problem s solved, n a second step, capacty margns are added to ts soluton. Assumptons: - flght arrvals occur accordng to nomnal schedules, - the ground handlng actvtes of all arrvng or departng arcraft are only performed at parng gates, - the ground handlng actvtes follow the same sequences for every consdered arcraft - the GHC nows the techncal characterstcs of the dfferent equpment and fleets, - the GHC has reference values for travel tmes and elementary ground handlng delays, - It s assumed by the GHC that all routes for each type of vehcle start and end at the correspondng base. - Each ground handlng unt can only perform one tas at one tme Adopted notatons Let us defne the consdered varables and parameters: - K :the set of arcraft nvolved n ground handlng actvtes durng the consdered tme perod ( K s ther number). - N : the number of dfferent servce fleets nvolved n ground handlng. F n,,2,..., : the amount of avalable vehcles of type. - - N F m,,2,..., arcraft stands. N F P p p,..., p -, 2 tot,: number of tass that a vehcle type can execute successvely at :set of avalable arcraft stand ( - H H h h,..., h : set of dfferent types of arcraft wth, 2 tot - ptot s ther number). h s ther total number). h j : tas duraton, the tme delay t taes to perform tas j on arcraft type h usng vehcle type. ( tot 06

127 Chapter 5 Decson Mang Processes for the Proposed Global Approach - - A d : scheduled start tme of ground handlng operatons for arcraft at ts parng stand. D d stand. : scheduled end tme of ground handlng actvtes for arcraft at ts parng - T :departure date of the arcraft from parng stand scheduled at D d. - l L : length of route l travelled by a vehcle of type. - S :parng stand of arcraft. S P. - - K : set of arcraft n competton wth arcraft to use vehcle type. pq : average travel tme,.e. the tme t taes to drve wth vehcle of type from arcraft parng stand p and to arcraft parng stand q. - p : average travel tme t taes to drve from the arcraft parng stand p to the base of the vehcle of type wth p P and,2,...,. N F - - V :average speed of vehcle type. C j : Start date of tas j performed by a vehcle of type on arcraft. The tass to load and unload luggage are supposed here to be performed by the same type of vehcle. Then for routes wth vehcle type : l 0 Z j ' f the route l type carres out the tas j on arcraft after coverng unloadng luggage tas on arcraft l Z j ' ' and 0 otherwse. f the vehcle number type l performs the tas j on the plane after completng loadng luggage onto arcraft l Z j ' and 0 otherwse. ' f the route number l wth vehcle type covers tas j on arcraft rght after performng t on arcraft ' and 0 otherwse for 2,3,4, Tentatve problem formulaton The above assumptons led to the formulaton of a nomnal overall optmzaton problem. Here the adopted objectve functon consders the mnmzaton of a convex mx of 07

128 Chapter 5 Decson Mang Processes for the Proposed Global Approach the sum of the arcraft departure delays and of the total dstance traveled by ground handlng vehcles: D d K NF n l mn l L (5.) T where 0 Constrants (5.2) and (5.3), shown below verfy that each ground handlng tas s assgned to a sngle route. lr l r 0 ' K j' n Z j, 2 ; K (5.2) n l l ' K Z ' 2,3,4, 5 ; K (5.3) Constrants (5.4) and (5.5) are route contnuty constrants: each vehcle after executng the tas assgned to t s supposed to leave the parng stand. lr 0 ' Z r K j' r 0 '' K Z lr j'' l l K Z K Z ' j '' ' j' ',2 l,2,..., n ; j, 2; K (5.4) l,...,n ; 2,3,4, 5; K (5.5) The set of nequaltes presented below descrbes the precedence constrants of operatons and the avalablty dates of servce vehcles. Indeed, for the constrant (5.6), the frst nequalty guarantees that a gven tas performed by a specfc vehcle on a gven arcraft cannot start before the prevous tas carred out by ths same vehcle has completely been performed on a prevous arcraft and the vehcle has travelled between the two parng stands and the second nequalty, specfes that a tas followng another one cannot start before the end of ths prevous tas (n ths case, t s the arrval of the arcraft to the parng stand). C C C d 2 A n l r 0 S C. Z lr 0 r n ' l r 0 ' K r r S ' S r. SS '. Z lr ' K (5.6) l 0 l In ths case, t s mposed that Z Z. 0 0 The operaton of dsembarng passengers does not requre the nterventon of a servce vehcle and can be carred out after the arrval of the arcraft to the parng area, so the only constrants to be consdered are: C d 6 A K (5.7) 08

129 Chapter 5 Decson Mang Processes for the Proposed Global Approach 09 The constrants related to the remanng operatons are establshed smlarly. For the santaton process, we get: K Z C d C Z C C n l l S A n l K l S S ' ' 3 ' 3 ' (5.8) Regardng the cleanng operaton, we have: C C K (5.9) For the caterng operaton, constrants are wrtten as: K Z C C C Z C C n l l S n l K l S S ' ' 2 ' 2 ' (5.0) As for the water process, constrants are: K Z C C C Z C C n l l S n l K l S S ' ' 4 ' 4 ' (5.) Wth respect to refuelng, the constrants are wrtten as: C C 6 3 K (5.2) Then, for loadng baggage: K Z C C C Z r r C C n l lr r S n l lr r K S S S S r r ' ' ' 2 0 ' ' (5.3) Here, also t was supposed that: l l Z Z

130 Chapter 5 Decson Mang Processes for the Proposed Global Approach C C C C K (5.4) and for the push bac operaton, we wrte: C C C C C n4 l ' K C C C n 5 l 2 5 S Z 4 C 5l 0 5 ' 5 S S ' Z 5l ' K (5.5) The departure of arcraft from ts parng stand can only be started after the completon of the push bac operaton (nequalty (5.6)) and t s not performed before the planned departure tme (nequalty (5.7)). T C K (5.6) 5 5 D T d K (5.7) At begnnng and endng of operatons for vehcle type, we have the constrants: 2 lr Z K r0 j j 2 lr Z K r0 j j For vehcles types 2, 3, 4 and 5 these constrants are wrtten: l K Z j l K Z j l,2,..., n (5.8) 0 l,2,..., n (5.9),2 ; 2,3,4,5 l,...,n,2 ; 2,3,4,5 l,...,n The travelled dstances by servce vehcles are gven by: (5.20) (5.2) 0

131 Chapter 5 Decson Mang Processes for the Proposed Global Approach L lr Z 0 0. K j r j S 2 lr Z 0 0. K j r j S lr r r.. Z V l ' 0 ' '. K K r S S SS ' L l K K K S. Z S l 0 ' K. Z l Z. V l 0 S S ' ' Analyss and soluton process,2 2,3,4,5 l,...,n l,2,..., n (5.22) (5.23) The optmzaton problem developed above s a mxed nteger problem. Varables * Z * are Boolean decson varables and varables C j and T are postve real decson varables. The frst varables correspond to the coverng of arcraft ground handlng needs by servce routes and the second varables correspond to the tme schedulng of actvtes along the servce routes. Each servce route s a duty to be performed by a correspondng servce team composed of a servce vehcle and a servce team. The sze of the problem s gven by: the number of decson varables composed of K 4n n and 7 K postve real varables, 2 Boolean varables - the number of nequalty constrants composed of n n constrants and 6 K nonlnear constrants the number of lnear equalty constrants: ( 2n 2 n 6 2n n K Note that each nonlnear nequalty constrant: z n m x y x 0, j, y j 0, Y j Where Y s an upper bound of y j j, s equvalent to: j ). K lnear, z R (5.24)

132 Chapter 5 Decson Mang Processes for the Proposed Global Approach For all x 0,, y 0, are satsfed: z x y n, j m, j Y 0, x, y 0,, j Y j z R and z R, where z x. y j f, and only f the constrants below z x z y z y z 0 j j n n, Y j. x j m n, n, j m j m (5.25) the nonlnear nequalty constrants (5.24) can be replaced only by the 3 th lnear nequalty of the system (5.25) [Bllonnet, 2007]. Then the whole optmzaton problem becomes a mxed nteger lnear problem whch can theoretcally be solved usng technques such as the Branch-and Bound algorthm [Land and al, 960].Clearly, ths approach even for small nstances of the problem (e.g. K 0 arcraft), t leads to a sgnfcant computaton tme when searchng for the exact soluton, for example usng a solver such as LP-Solve or CPLEX Numercal applcaton For example, a case wth 5 arcraft nvolved n 0 flghts wth 3 dfferent ground handlng operators performng 4 dfferent ground handlng actvtes, has been consdered numercally. In ths case, the objectve functon to mnmze reduced to the sum of the delays whch are generated by the assgnments of the ground handlng unts to the ground handlng tass. In fgure 5. s represented the structure and duraton assumed for the ground handlng actvtes. Then table 5.2 provdes the nomnal arrval and departure schedules as well as the assgned parng postons. 2

133 Chapter 5 Decson Mang Processes for the Proposed Global Approach 2. Refuellng (9 mnutes) 0. Arrval. De-boardng (7 mnutes) 4. Boardng (5 mnutes) 5. Departure 3. Caterng (0 mnutes) Fgure 5. : Structure and duraton of the ground handlng actvtes Arcraft Scheduled Arrval Tme Scheduled Departure Tme Parng Poston Table5. : The nomnal arrval and departure schedules and the parng postons of arcraft Ths problem has been solved usng the lbrary LP-Solve whch has been run on a personal computer. An exact assgnment of the ground handlng unts to the ground handlng tass has been computed. Table 5.2 provdes the correspondng assgnment soluton whch s graphcally represented n fgure 5.2. Fgure 5.2 : Vehcles routes 3

134 Chapter 5 Decson Mang Processes for the Proposed Global Approach Arcraft Scheduled Start Tme Tas Tas Start Tme Tas Completon Tme Scheduled End Tme Table5.2 : The assgnment soluton The sum of the delays at departure for the arcraft accordng to ths soluton s equal to 38 mnutes whch tends to ndcate that ground handlng resources were n ths case nsuffcent to tacle effcently the nomnal arrval/departure schedule. The soluton for ths very small problem was obtaned after.37 mnutes of computaton. When consderng slghtly larger nstances of ths problem, the computaton tme ncreases very sharply to excessve values (tens of mnutes and soon, hours of computaton). Then ths exact soluton approach does not loo sutable to treat real sze assgnment problems (wth for nstance no less than varables and constrants for an nstance nvolvng 690 flghts. It s expected that ths stuaton wll reman even f specalzed versons of the resoluton software were developed or f a faster computer was employed. Ths consttutes a strong lmtaton for ths approach. So t appears of nterest to consder the development of a heurstc approach whch can be able to produce feasble solutons n a very short computaton tme. Ths wll allow the 4

135 Chapter 5 Decson Mang Processes for the Proposed Global Approach manager, here the GHC, to restart the soluton of ths problem when the current operatonal condtons become rather dfferent from the predcted ones The proposed GHC heurstc Let us consder durng a perod of operatons, wth a set K of arrvng and departng arcraft to/from the stands. Here we develop a greedy centralzed heurstc whch wll ensure the feasblty of all ground handlng operatons. The dea of the centralzed heurstc s to ran arrvng and departng arcraft accordng to ther planned start tme of the correspondng ground operatons (ether arrval ground handlng tass or departure grand handlng tass). Then the central planner wll process n ths order each arcraft ground handlng actvty by lnng each tas to a route to buld a ground handlng duty: - To cover tas j at arcraft t wll search between the already created routes of type j, whch one can cope wth t, wthn the planned nterval and at lower transportaton cost. - If none of the exstng route provdes a feasble soluton. and there are remanng capacty of type j at the correspondng base, a new route of type j startng at ths base s created wth frst stop at arcraft. 2. and there are no remanng transport capacty at base of type j, add ths tas at the route of type j whch mnmzes the mx of resultng delay for arcraft and of dstance travelled to reach t wth the weght. Then repeat wth all the expected ground handlng tass j at an arrvng or departng arcraft. Ths wll produce feasble sets of dutes (routes) to be performed by the dfferent ground handlng fleets and worforce. Then ths data wll be used by the ground handlng coordnator to compute, accordng to the process proposed n the prevous chapter, the level of resources that each ground handlng manager must provde at each tme perod. These resources wll be afterwards ether effectvely used to process arcraft and passengers or wll reman as a warm reserve to face perturbatons and ncdents. 5

136 Chapter 5 Decson Mang Processes for the Proposed Global Approach 5.3. Decentralzed fleet management Classes of fleet management problems The fleet management problems consdered here correspond to the parng problems that have to be solved by the ground handlng managers between planned demand of specalzed ground handlng servces and the correspondng avalable ground handlng resources. Tang nto account that some servce provders must perform two dfferent tass, t appears necessary to separate ground provder fleet servces nto two categores: the frst, C, ncludes the provders who perform two dfferent and non-consecutve tass as: the servce provders who tae care of both the loadng and unloadng luggage, and the servce provders who tae care of both the boardng and de-boardng of passengers. The second category, C, gathers the 2 provders who carry a sngle type of tas ether on an arrvng or departng arcraft Adopted notatons The formulatons of the consdered to classes of fleet management problems adopt the followng notatons: Each tas of the turnaround process t,...,ts carred out on an arcraft,...,i t by a specfc servce provder t,...,k ; Precedence constrants descrbe executon orders for pars of tass; I t :s the set of all arcraft that requre servce from the ground provder durng a perod of t t tme; I t s the set of arcraft that have requred servce n the recent past; I the set of arcraft p f that wll requre servce n the near future; I t I t p I t f Each servce provder operates a fleet of homogeneous vehcles; x,..., X t x t a, j equal f vehcle, x, x,..., X t whch performed the tas t, t,...,t j, j,..., I t, mmedately after servng arcraft,,..., I t serves arcraft Each arcraft, t A I, has a scheduled arrval tme d and a scheduled departure tme D d ; t Each tas t has a release tme b from whch t can be started and a completon tme the tme at whch the arcraft, t I, s expected to request servce. 6 t f. b s t j

137 Chapter 5 Decson Mang Processes for the Proposed Global Approach Each tas t has a non-preemptve processng duraton S t ; D, s the dstance to drve from an arcraft parng stand and to an arcraft parng stand j; j T s the set of tas that wll be performed on the arcraft once the agent t t completes the executon of ts tas t; that wll be carred out by the agent t ; T s the set of tas that were performed on the arcraft before the tas t Formulaton of the GHFAS problem (C case) The optmzaton objectve s a mx of the sum of generated delay at the unloadng stages and at the loadng stages wth the total travelled dstance by the correspondng fleet. t D t ' t D t ' mn f j d j t ' T S j 2 f ' j d j t ' T S j xx I t ji t r r. D rd D 2 p f x, t, r 0, j, t t, j, j where 0, 2 0, 2 under the followng constrants ncludng the assgnment coverng constrants: a (5.26) x, t, r t 0 r xx I t p a j I, j f, t T (5.27) x, t, r 0 r I ji t p a f, j t t x X, t T (5.28) x, t, r 0 t a r I, j r 0 I t p f a x, t, r, j t j I, x X, t T (5.29) t f,, a x t r, j 0, t t t I p, j I f, x X, t T, r 0, (5.30) b d t j A j t j I f, t T (5.3) b t j r b t S t, r D V, j x t t, r j t r, t t, r b S a x,, ' D D V x, j t j I, x X, r 0,, t f t T (5.32) f b S j t, 0 t t,0 j j t j I f, t T (5.33) t b j S d t, 0 D t ' j j t' T t j S j I f, t T (5.34) 7

138 Chapter 5 Decson Mang Processes for the Proposed Global Approach t b' j b t j S t,0 j T t j I f, t T (5.35) b' t j r b t S t, r D V, j x t t, r j t r, t t, r b S a x,, ' D D V x, j t t j I f, x X, t T, r 0, (5.36) t b' j S t, j d D j t ' T S t ' j t j I f, t T (5.37) f t, j t b' j S t, j t j I f, t T (5.38) b' E t j j j I (5.39) t f Here the decson varables are relatve to the assgnment of vehcles to arcraft (Boolean) and the scheduled start tme of each elementary ground handlng tas (real) Formulaton of the GHFAS problem (C2 case) For each sngle tas ground handlng fleet we get the followng formulaton of the GHFAS problem: mn t D t' t xx I ji x, f d t T S t ' t D a j j j where s a postve parameter and wth the followng constrants: x, t xx I t p a, j t j I f, t T2 p f, j, j (5.40) (5.4) t t x, t a I p j I f, j t x X, t T2 (5.42) x, t a a I t t p, j I f x, t j, t t j I f, x X, t T (5.43) 2 a x, t, j 0, t t t I p, j I f, x X, t T (5.44) 2 b d t j A j t j I f (5.45) b t j b t S t D V, j x a x, t, j t t j I f, x X, t T (5.46) 2 t t f j b j S t j t j I f, t T2 (5.47) 8

139 Chapter 5 Decson Mang Processes for the Proposed Global Approach t t D t' t b j S j d j S t ' T j j I f, t T (5.48) 2 The equaton (5.27) and (5.4) ensure all arcraft receve servce. Equatons (5.28) and (5.42) mpose that all the vehcles can begn and end ther servce tour at any poston. Equaton (5.29) and (5.43) are flow conservaton constrants: a vehcle arrvng at an arcraft must leave that arcraft later. Equatons (5.30) and (5.44) ensure each possble tas s ether assgned or not. The nequalty (5.32), (5.36) and (5.46) provde earlest start tme constrants for the servce at a ready arcraft tang nto account the travellng tme between arcraft. The nequalty (5.3), (5.39), (5.45)specfy that a tas followng another one cannot start before the end of ths prevous tas (precedence constrants). The nequalty (5.34), (5.37) and (5.48) defne the latest start tme for each servce tang nto account the actvtes that would be performed after. The equatons (5.33), (5.38) and (5.47) represent the endng tme of each tas consderng the startng tme whch has been already computed and the tas duraton On lne Ground Handlng Fleet Assgnment (GHFA) problem at the level of each GHM Ground Handlng Fleet Coordnaton To perform the ground handlng actvtes for each arcraft wthn the allocated tme, these dfferent ground handlng fleet servces have to coordnate between each other whle respectng the constrants of schedulng tass for each arcraft and the constrants related to the use of ground handlng unt: equpment, manpower, vehcle, etc accordng to the he organzaton presented n the Chapter Proposed heurstcs for on-lne GHFA In a nomnal stuaton, the ground handler fleet managers wll assgn a vehcle and a wor team to each route. Ths vehcle may be changed by another to pursue the duty n accordance wth operatonal consderatons (refuelng need, mechancal falure, etc) whle wor teams wll be shfted accordng to labor and safety regulatons. 9

140 Chapter 5 Decson Mang Processes for the Proposed Global Approach Here t s supposed that there are enough spare vehcles and wor teams to meet operatonal perturbatons: When an arrvng arcraft s delayed whle hs predcted arrval tme s avalable, the ground handler feet manager can tae, ndependently of the other ground handlng fleet managers, one of the three followng decsons: - mantan the correspondng ground handlng tas n the duty at the same place n the sequence. In that case the resultng delays should be ntegrated nto the schedulng of the duty. - mantan the correspondng ground handlng tas n the duty but at another place n the sequence. - delete the correspondng ground handlng tas from the duty and assgn t to another duty or to a spare vehcle and team (local duty) to perform the tas when the arcraft wll be avalable. When a departng arcraft s delayed for some external reason (arport, arlne, ATC), one of the three followng decsons must be taen by each ground handler fleet manager: - mantan the correspondng ground handlng tas n the duty at the same place n the sequence. In that case the resultng delays should be ntegrated nto the schedulng of the duty. - mantan the correspondng ground handlng tas n the duty but at another place n the sequence. - delete the correspondng ground handlng tas from the duty and assgn t to another duty or to a spare vehcle and team (local duty) to perform the tas when the arcraft wll be avalable to start departure ground handlng actvtes. From the solutons of the assgnment problems solved by each ground handlng manager, the ground handlng coordnator forward the mlestones correspondng to the completon of ground handlng actvtes to the arlnes and the ATC to produce f necessary new estmates for the departure schedule of the arcraft. 20

141 Chapter 5 Decson Mang Processes for the Proposed Global Approach 5.5. Case study Arport and ground handlng characterstcs To the best of our nowledge, no benchmar nstances exst for ths problem. Then, a real traffc data from Palma de Mallorca Arport has been consdered. Palma de Mallorca Arport s, wth respect to arcraft and passengers traffc, the thrd largest Spansh arport. Durng the summer perod t s one of the busest arports n Europe, wth 22.7 mllon of passengers n 20. The arport s the man base for the Spansh carrer Ar Europa and also a focus arport for German carrer Ar Berln. It occupes an area of 6.3 m2 (2.4 sq m). Due to rapd growth of arcraft traffc and passenger flows along the last decades, addtonal nfrastructures have been added to the two orgnal termnals A (bult n 965) and B (bult n 972). Palma de Mallorca Arport s composed now of two runways, four termnals and 80 parng stands wth 27 of them at aprons It can handle up to 25 mllon passengers per year, wth a capacty to dspatch 2,000 passengers per hour [PDM, 202]. Fgure 4 dsplays the hourly traffc of arrvng and departng arcraft on a typcal summer day at ths arport. It appears that arcraft traffc remans ntense from early mornng untl the begnnng of nght hours. 40 Arrval arcaft 's number 35 Departure arcraft 's number Fgure 5.3 : 0 /08/2007 Palma de Mallorca Arport Arcraft hourly traffc The followng datasets were used n order to create the nstances: a) One day flght traffc data from the Palma de Mallorca arport correspondng to a summer busness day (345 arrvals of arcraft and 345 departures of arcraft) was consdered. Ths ncludes the lst of the arcraft performng a turnaround durng the 2

142 Chapter 5 Decson Mang Processes for the Proposed Global Approach day, the scheduled arrval and departure tmes, the real arrval and departure tmes, the type of arcraft, and the parng poston. b) Dstances between the parng postons and between them and the depot. The Palma de Mallorca arport has 80 parng stand: 27 of them are remote stands. A constant velocty was used to calculate the vehcle travelng tme. c) Tass nformaton: usng the specfcatons of the arcraft manufacturers (Arbus, 2005; Boeng 200, ATR 999), three types of arcraft wth dfferent szes were modeled. For each operaton ncluded n the problem and accordng to the type of arcraft, the duraton, the precedence restrctons regardng the other tass, and the type of vehcle used have been consdered Implementng the global plannng of ground handlng resources The developed heurstcs have been mplemented n Java. As t has been mentoned on the chapter 4, ths approach s proposed to calculate the nomnal number of resources requred for each ground handlng manager durng a day of traffc. The heurstc proposed s a greedy heurstc. The soluton of ths approach s gven n the Table 5.3. It represents the number of the arcraft whch wll be performed by each ground handlng unt of each ground handlng servce provder. Ground handlng GHU GHU 2 GHU 3 GHU 4 GHU 5 GHU 6 GHU 7 GHU 8 GHU 9 actvty De-boardng/ Boardng passengers Unloadng/ Loadng baggage Caterng Cleanng Refuellng Santaton Potable Water Supply Push bac Table5. 3 : Soluton of herarchcal approach 22

143 Chapter 5 Decson Mang Processes for the Proposed Global Approach Usng ths soluton, only 2 arcraft wll have a delay at the level of the departure tmes wth a maxmum delay of 4 mnutes. The 4 arcraft that would leave ther parng stand later that whch t had been predcted ther departure tmes match wth busest flght traffc perod. Fgure 5.4 represents the hourly dstrbuton of arcraft the departure delays resultng from the proposed heurstc. Nuber of arcraft departure delay 3,5 3 2,5 2,5 0,5 0 Tme Fgure 5. 4 : Hourly delays dstrbuton resultng from the proposed heurstc The proposed global plannng heurstcs of ground handlng resources has been calculated usng the dataset presented n the precedent paragraph. Ths global plannng of ground handlng resources as t has been descrbed n the chapter 4 s composed of three steps. For the frst step, t has been supposed that the nomnal number of each ground handlng resources s presented n the fgure number of vehcle Deboar dng\ Catern g 's Clean ng 's Unload ng\l Fuelln g 's Santat on 's Water 's fleet Pushn g Fgure 5.5 : Nomnal composton of ground handlng fleets In the second step, the unt tme perod whch has been consdered has been taen equal to the maxmum between 5 mnutes and the smallest duraton of a ground handlng operaton, ncludng transfer tme: 23

144 Chapter 5 Decson Mang Processes for the Proposed Global Approach jk t j u t max 5, mn s (5.49) Ground handlng actvty Duraton (mn) De-boardng passengers 5 Caterng 5 Cleanng 5 Boardng passengers 5 Unloadng baggage 5 Fuellng 5 Loadng baggage 5 Santaton 5 Potable water supply 5 Push-bac 5 Table5.4:The unt tme perod of each ground handlng operaton results The thrd step of the estmaton of the necessary resources at a gven tme for all ground handlng managers s performed by addng margns to the nomnal level of demand of scheduled arrval and departure flghts. Ths s done accordng to formula (4.5), (4.6) and (4.7). The fgures presented below provde the sze of the resources requred for each ground handlng manager to perform ther correspondng ground handlng tass n case of perturbatons that can occur durng the day. As t can be seen, the number of reserved resources ncreases n the busest flght traffc perod (arrval/departure arcraft) accordng to the fgure 5.6. Number of resources Unloadng /Loadng baggage 00:00 0:45 03:30 05:5 07:00 08:45 0:30 2:5 4:00 5:45 7:30 9:5 2:00 22:45 Tme Number of resources De-boardng/Boardng passengers :00 0:45 03:30 05:5 07:00 08:45 0:30 2:5 4:00 5:45 7:30 9:5 2:00 22:45 Tme 24

145 Chapter 5 Decson Mang Processes for the Proposed Global Approach Number of resources Santaton 00:00 0:45 03:30 05:5 07:00 08:45 0:30 2:5 4:00 5:45 7:30 9:5 2:00 22:45 Tme Number of resources Refuelng 00:00 0:45 03:30 05:5 07:00 08:45 0:30 2:5 4:00 5:45 7:30 9:5 2:00 22:45 Tme Number of resoureces Cleanng 00:00 0:45 03:30 05:5 07:00 08:45 0:30 2:5 4:00 5:45 7:30 9:5 2:00 22:45 Tme Number of resources Caterng 00:00 0:45 03:30 05:5 07:00 08:45 0:30 2:5 4:00 5:45 7:30 9:5 2:00 22:45 Tme Number of resources Push bac 00:00 0:50 03:40 05:30 07:20 09:0 :00 2:50 4:40 6:30 8:20 20:0 22:00 Tme Number of resources Potable water supply 00:00 0:50 03:40 05:30 07:20 09:0 :00 2:50 4:40 6:30 8:20 20:0 22:00 Tme Fgure 5.6: Number of the resources requred for each ground handlng actvtes each of perod of tme Implementng the heurstcs for on-lne GHFA To test the effcency of ths approach, the accurate arrval tmes of each consdered flghts are supposed to be communcated to the ground handlng managers thrty mnutes before the effectve landng. Here, ths allows the ground handlng managers to reassgn the ground handlng resources by consderng the updated arrval tmes at the parng stands of 25

146 Chapter 5 Decson Mang Processes for the Proposed Global Approach the flghts announced to land wthn the next half hour. Arcraft wthn fve mnutes to land have been supposed to mantan the prevous assgnment soluton. No flght drected towards the consdered arport has duraton less than forty mnutes. Then the real departure tmes where compared wth the ones obtaned through the proposed heurstc approach. The consdered ground handlng resources were the ones effectvely exstng at that arport. The applcaton of the proposed heurstc approach to the nomnal schedule of arrvals durng the consdered reference day provded a feasble assgnment for each ground handlng manager n at most 0.3 seconds. These solutons led to delays wth respect to scheduled departure schedule nvolvng only 36 arcraft, wth a maxmum delay of 6 mnutes. The average delay among delayed arcraft has been of 7 mnutes. Fgure 5.7 dsplays the hourly dstrbuton of delayed arcraft at departure resultng from the applcaton of the proposed decentralzed approach. Clearly, the occurrence of these delays corresponds to the busest arcraft traffc perods at the arport where ground handlng resources become short. The proposed heurstc could be restarted usng hgher ground handlng resource levels provded by the ground handlng coordnator to mprove the expected delay performance of the system :00 02:00 04:00 06:00 08:00 0:00 2:00 4:00 6:00 8:00 20:00 22:00 00:00 Fgure 5. 7 : Hourly delays dstrbuton resultng from the proposed heurstc Hstorcal data from 0/08/2007 at Palma de Mallorca Arport ndcate that about 244 arcraft departures where delayed for multple reasons, ncludng one of the man reasons, ground handlng delays. The maxmum observed delay s about 520 mnutes and the average delay among delayed arcraft has been of 30 mnutes. There s nformaton about the use of a partcular system to manage ground handlng at that arport. 26

147 Chapter 5 Decson Mang Processes for the Proposed Global Approach It s clear, that n theory, the proposed heurstc approach provde sgnfcantly mproved results wth respect to departure delays. Then t can be expected for ths partcular arport that, even f the mplementaton of the proposed heurstc approach s not perfectly performed, some notceable mprovement wth respect to the current practce wll tae effect. Ths s qute noteworthy snce the proposed heurstc has not been partcularly mproved wth respect to a basc greedy approach Concluson In ths chapter the soluton of the dfferent assgnment problems solved by the ground handlng coordnator and ground handlng managers has been consdered. An exact approach has been adopted at frst to solve the global assgnment problem consdered n the proposed framewor by the ground handlng coordnator. Numercal results usng LP-Solve show that beyond the case of very small problems (0 to 2 flghts), the exact approach s not able to produce the optmal soluton n an acceptable tme. So a greedy heurstc has been developed n that case. In the case of the parng problems faced by the ground handlng managers, even f the correspondng optmzaton problems are of smaller sze that the one faced by the ground handlng coordnator, only the heurstc approach has been developed. The whole process has been llustrated by consderng a case study wth real traffc where t has been assumed that flght arrval tmes are perfectly nown half an hour n advance. Even f scheduled and effectve arrval tmes are dfferent, the adopted traffc stuaton can be consdered as normal. In the next chapter, the proposed framewor for ground handlng management wll be dscussed n the case of huge traffc perturbatons characterzng an arport dsrupton. 27

148 Chapter 5 Decson Mang Processes for the Proposed Global Approach 28

149 Chapter 6 Ground Handlng Management Under Dsrupton 6. CHAPTER 6 GROUND HANDLING MANAGEMENT UNDER DISRUPTION 29

150 Chapter 6 Ground Handlng Management Under Dsrupton 30

151 Chapter 6 Ground Handlng Management Under Dsrupton 6.. Introducton In ths chapter s consdered the case n whch an arport s subject to a large perturbaton whch n general affects all ts sub-systems: runway operaton, arsde taxng operaton, ground handlng operatons, passenger termnals and groundsde land traffc. Ths drastc stuaton termed arport dsrupton has been defned n qualtatve terms and very few specfc studes to cope systematcally wth t are avalable. In ths chapter, after tryng to better dentfy ths stuaton, a new formalsm s ntroduced to cope wth the uncertanty assocated to the duraton of many actvtes n ths stuaton. Then a tentatve approach to desgn a decson process for the ground handlng coordnator to better cope wth ths stuaton s proposed. Ths adapted decson process s based on the assessment of the crtcalty of each arrvng or departng arcraft n the reducton of the dsrupton stuaton, rrespectve of drect ground handlng operatons costs Arport Dsrupton Defnton of arport dsrupton To our nowledge there exsts no specfc defnton for arport dsrupton whle some recent wors refer to ths stuaton [Ploog, 2005] and [Tanger and al, 203] wthout provdng any defnton. Accordng to the Brtsh Standards Insttute [Busness contnuty management, 2006], a dsrupton s an event whch causes an unplanned, negatve devaton from the expected delvery accordng to the organzaton s objectves. Accordng to ths defnton, the term dsrupton could be perceved as equvalent to the term perturbaton. The ground handlng servces are delvered n a changng envronment wth many operatonal uncertantes. For example, the expected arrval tmes for flghts are subject to frequent delays, the duraton of ground handlng tass s senstve to unexpected events such as addtonal travel tme due to traffc congeston on arsde servce ways or machne breadowns. Then t could be consdered that ground handlng management tacles n permanence dsrupted stuatons. In the Ar Transport management lterature, the ssue of arlne dsrupton management has been consdered more early [Kohl and al, 2007], [Clausen and al, 2005] and 3

152 Chapter 6 Ground Handlng Management Under Dsrupton has been assocated wth the arlnes recovery problem [Batu and al, 2006], [Lettovsy and al, 997]. In fact, for these authors a dsrupted stuaton occurs when a successon of unexpected events leads the system state out of range of the current operaton practce whch s no more able to compensate devatons and mae the system state to return near a nomnal stuaton. In that case, recovery actons must be taen to avod a cumulatve degradaton of the performance of the system. In ths chapter, ths later understandng of a dsrupted stuaton wll be transposed to the case of arport management where dsrupton management should also cope wth some crss stuatons Consequences of arport dsrupton Here the operatonal stuaton whch s consdered s the one n whch, as a consequence of some event or successon of events, the whole arport operaton s perturbed and presents at the same tme mportant delays and large uncertantes wth respect to effectve arrval and departure tmes. Possble consequences of an arport dsrupton stuaton can be [Ploog, 2005]: - for passengers: canceled departng flghts or loss of connecton flghts by passengers (delayed arrval at stand of prevous flght, delayed transfer of passengers and luggage towards the followng flght), passengers who are oblged to wat for long perods wthout precse nformaton at boardng gates or n the arcraft once boarded. - for crews: mpossblty for a crew member to contnue ts scheduled flght parng, dffcultes for arlnes to consttute techncal and commercal crews for departng flghts. - for arcraft: unavalablty of an arcraft to perform a scheduled departng flght, dffculty to perform scheduled sde actvtes such as mantenance actvtes Sources of arport dsrupton Causes for the arport dsrupton stuaton can be related wth ncomng traffc, the arport tself and exogenous events. 32

153 Chapter 6 Ground Handlng Management Under Dsrupton Wth respect to ncomng traffc, arport dsrupton can be generated when a large share of the ncomng traffc durng a perod of tme, for example a pea hour for the arport, arrves late wth large delays. Ths can be the result of bad weather condtons, of a temporary lac of capacty of the ar traffc system caused by an excess of traffc demand, or by the reducton of effectve ATC capacty as a result of some socal or techncal problem. Whle the ATFM system [Gwggner, 2004] maes the excess of demand stuaton very unlely, the ATC system presents n general hgh levels of relablty and avalablty. Wth respect to the arport tself, arport dsrupton stuatons can be produced by a temporary lac of capacty caused for example by the closure of a runway, bad weather condtons (fog, snow, strong ran), the lac of suffcent ground nstallatons and equpment to cope wth a pea of traffc, socal problems (stre of some category of arport employees), occurrence of hazards at the arport (crash of landng or departng arcraft, huge fre). Exogenous causes whch can result n arport dsrupton are transent stuatons assocated to the recovery from the effect of natural hazards (volcano ashes, tsunam, nuclear alerts) or from overfly restrctons n conflctve areas Ground Handlng Management Objectves and Operaton under Arport Dsrupton Here t s consdered that the management of ground handlng durng an arport dsrupton should contrbute to ts reducton and elmnaton. Ths mples eventually the defnton of new objectves and new decson processes to be adopted durng ths transent stuaton. In such a stuaton, t can be expected that the proposed decentralzed ground handlng management should be more strongly drven by the ground handlng coordnator to tacle wth prorty the overall arport objectves Ground handlng management objectves under arport dsrupton In ths stuaton, the whole operatons plannng performed by ground handlers must be revsed wth temporary new objectves: 33

154 Chapter 6 Ground Handlng Management Under Dsrupton - Contrbute to the return of arport operatons to a near nomnal stuaton as soon as possble snce the dsrupted stuatons reduce the overall arport performance and servce offered to the passengers. Ths can be done through the adopton of more costly ground handlng solutons. - Lmt as much as possble the maxmum flght delays nstead of the mean passenger delay adopted n regular arport operatons. - Mnmze the number of mssed passenger connectons. Ths has an mportant contrbuton onto the performance of the arport. In general, the most of passenger mssed ther connecton because of ether the ground handlng operators whch they dd not taen nto account the mpact of delayng the performng the ground handlng actvtes of ths flght or of the bad manner of sharng nformaton between the A-CDM partners A proposal for ground handlng management under arport dsrupton Here t s proposed, wth the objectve to handle the overall arport objectves, at the ground handlng coordnator taes over the drecton of the ground handlng management by mposng to the ground handlng managers, prorty lsts of flghts to be processed. The reorderng of the scheduled arrvals and departures nto prorty lsts wth respect to ground handlng by the ground handlng coordnator can be the result of: - a negotaton wth the other A-CDM partners about specal demands from them, - the assessment of the current and near future ground handlng stuaton accordng to current and predcted traffc of arcraft, - the occurrence of some ground handlng ncdent (equpment falure). The ground handlng coordnator wll provde onlne to the ground handlng managers two frequently updated prorty lsts: - one s relatve to arrvng arcraft, - the other one s relatve to departng arcraft. 34

155 Chapter 6 Ground Handlng Management Under Dsrupton An arrvng arcraft wll enter these two prorty lsts when tspredcted arrval tme at the parng stand becomes smaller than the ground management operatonal horzon. An arrvng or departng arcraft wll leave the correspondng lst when ts ground handlng processng s ready to start. An arcraft can be at the same tme n these two prorty lsts, so these lsts are not ndependent. Here, ground handlng resources are also separated between those whch are dedcated to arrvng arcraft and those whch are dedcated to departng arcraft. Then ground handlng managers wll assgn ther respectve resources accordng to these prorty lsts. Ths wll mae that many arrvng or departng arcraft wll not be necessarly processed accordng to ther ran n the arrvng or departng tme schedules. Snce n ths stuaton demand levels may overpass avalable ground handlng capacty, the ground handlng coordnator establshes these prorty lsts for ground handlng managers wth the objectve to reduce or avod cumulatve effects whch wll otherwse contrbute to prolong the dsrupted stuaton of the arport. In ths case, tang nto account the uncertanty about the completon of many events at the arport arsde, the ground handlng coordnator wll requre from some ground handlng managers to put nto alert all ther effectve ground handlng resources. For example ths could be the case wth the de-cng capacty of an arport. For others ground handlng actvtes, the ground handlng coordnator can adopt a tme-of-the-day polcy based on pre computed reserves to mae ready ground handlng extra resources. In that case, t s consdered that the pool of ground handlng resources necessary to perform arrvng or departng ground handlng actvtes are requred to be avalable at the parng place as soon as possble and start ther actvtes accordng to the ground handlng sequence assocated to ths arcraft. For example, one of the objectves wth respect to flght arrvals s to mnmze the watng tme for de-boardng passengers and luggage, another one s to mae sure that passengers embar n the arcraft wth a mnmum delay, f any, wth respect to the rescheduled flght departure tme. So, they wll be n charge of moblsng n due tme the necessary ground handlng resources for flght arrval or flght departure processng. Arport ar traffc control servces update the predcted arrval tmes whch are forwarded to arport servces, ncludng arlnes and ground handlng. Ths starts the process of updatng 35

156 Chapter 6 Ground Handlng Management Under Dsrupton the assgnment and schedulng of tass for each ground handlng fleet. In the case n whch repeated arcraft arrval schedule perturbatons are occurred or are expected, accordng for nstance to meteorology condtons, the horzon of the dfferent ground handlng fleet management problems can be commonly lmted to no more than two hours ahead. Each ground handlng manager wll solve the new nstance of each GHFA problem by applyng some nd of the heurstc such as the one descrbed n the prevous chapter but modfed wth respect to one pont: Instead of treatng each flght accordng to ts poston n the arrval or departure schedules, each flght wll be treated accordng to ts updated prorty ran n the correspondng arrval or departure lst Operatonal uncertanty durng arport dsrupton In general n an arport dsrupton stuaton, whch s generated n general, as dscussed above, by a successon of unexpected perturbatons, many parts of the arport start behavng out of nomnal condtons generatng ncreased travel and servce tmes as well as a hgher dstrbuton of them. Although ground traffc s always performed n complance of prorty rules between vehcles of the same type and between vehcles of dfferent types along the dfferent ground tracs of the arport, multple queues of arcraft and ground servce vehcles may grow and nteract. To be reactve to the dsrupton stuaton, ground handlng resources must be ready to enter nto acton once a hgh prorty flght arrves at the parng stand or when a hgh prorty flght has to prepare for departure. Then, the ground handlng management should wor out decsons based on some predcton of arrvals or departures tmes from the parng stands and by adoptng some estmates for servce vehcle travel tmes as well as for ground handlng actvtes duratons. Consderng the hgh degree of uncertanty wth respect to tmng and delays, a determnstc approach, such as through determnstc optmzaton, to tacle ths stuaton appears neffectve [Rav and al, 2004]. On the other sde, the adopton of a probablstc approach wll be unfeasble by lac of statstcal data on one sde and by the resultng cumbersome computaton needs [Dyer and al, 2003]. Then, s the followng subparagraph, an ntermedate approach where uncertanty s dsplayed but treated through rough processes wll be proposed. In the consdered case, the ground handlng coordnator s 36

157 Chapter 6 Ground Handlng Management Under Dsrupton supposed to generate the prorty lsts accordng to the current and predcted ground handlng stuatons. These lsts, as t has been mentoned before, wll be provded on lne to the dfferent ground handlng managers who wll mae a copy of them. Fgure 6. descrbes the ground handlng management under dsrupton by generaton the prorty lsts at the level of the GHC. Fgure 6. : Ground handlng management under dsrupton Snce these prorty lsts can be modfed at the ground handlng coordnator level accordng to the occurrence of unexpected events, ths could mply that the assgnments of ground handlng unts to flghts should be changed n accordance. To provde some stablty to the assgnments performed by the ground handlng managers, t has consdered that once a ground handlng unt starts to turn ready to perform an actvty at a gven flght, ths assgnment s defntve and the correspondng flght s deleted from the lst of the 37

158 Chapter 6 Ground Handlng Management Under Dsrupton correspondng ground handlng manager. Ths wll happen only wth flghts whch are close to be processed. Fgure 6.2 represents the process of the ground handlng management under uncertanty at the two level of the proposed ground handlng management organsaton structure: GHC and GHMs. Operatonal Horzon GHC F F 2 F 3 F 4 F 5 F 6 F 7 F 8 GHM F F 2 F 3 F 4 F 5 F 6 F 7 F 8 GHU GHU 2 GHU GHU X Operatng made ready to perform ts next actvty Fgure 6.2 : Operatonal uncertanty durng arport dsrupton 6.4. Adopted representaton of uncertanty In the followng, to represent uncertanty wth respect to the tme occurrence of events or the duraton of actvtes, duratons wll be represented by fuzzy dual numbers [Cosenza and al, 20], [Cosenza and al, 202]. 38

159 Chapter 6 Ground Handlng Management Under Dsrupton Some elements about fuzzy dual numbers such as number. The set of fuzzy dual numbers s the set ~ of the dual numbers of the form a b a, b where a s the prmal part and b s the dual part of the fuzzy dual Observe that a crsp fuzzy dual number wll be such as b s equal to zero, loses both ts dual and ts fuzzy attrbutes. To each fuzzy dual number s attached a fuzzy symmetrcal number whose graphcal representaton s gven below where μ s a symmetrcal membershp functon defned over R: μ 0 a-b a a+b R Here we recall some basc operatons wth fuzzy dual numbers. The fuzzy dual addton of fuzzy dual numbers, wrtten ~, s dentcal to that of dual numbers and s gven by: x. y x. y x x y y (6.) Its neutral element s 0 0., wrtten ~ 0. The fuzzy dual product of two fuzzy dual numbers, wrtten, s gven by: Fgure 6.3 : Representaton of a fuzzy dual number x. y x. y x. x. x y x y (6.2) The fuzzy product has been chosen n that way to preserve the fuzzy nterpretaton of the dual part of the fuzzy dual numbers but t maes a dfference wth classcal dual calculus. The neutral element of fuzzy dual multplcaton s 0., wrtten ~ and only non-zero crsp numbers have an nverse. Both nternal operatons, fuzzy dual multplcaton, are commutatve and assocatve, whle the fuzzy dual multplcaton s dstrbutve wth respect to the fuzzy dual addton. Observe that the nlpotent property of operator s mantaned: 2 ~ 0. It appears also that fuzzy dual calculus s qute smpler than common fuzzy calculus ([Kosnsy, 2006], [Nasser, 2006])

160 Chapter 6 Ground Handlng Management Under Dsrupton 40 The pseudo ( ~ s not a vector space) norm of a dual fuzzy number s gven by:. b a b a (6.3) where 0 s a shape parameter. The shape parameter can be defned as: dy y b y. 2 (6.4) Fgure 6.3 dsplays standard fuzzy symmetrcal numbers wth dfferent shape parameters. Fgure 6.4 : Examples of shapes fro fuzzy dual numbers The followng propertes are met by ths pseudo norm whatever the values of the shape parameters: 0. : ~. b a b a (6.5) R b R a, 0 0. b a b a (6.6).... b a b a R b R a,,, (6.7) b a b a.... R b R a,, (6.8) Partal orders between fuzzy dual numbers can be ntroduced usng the above pseudo norm. Frst a strong partal order wrtten can be defned over ~ by: : ~.,. b a b a b a b a b a b a (6.9) Then a wea partal order wrtten can be also be defned over ~ by: : ~.,. a b a and b a b a b a b a b a b a (6.0) Fgures 6.4 and 6.5 dsplay dfferent partal orders between pars of dual fuzzy numbers and nequaltes between fuzzy dual numbers are qute dfferent from those used wth classcal fuzzy numbers b a b a

161 Chapter 6 Ground Handlng Management Under Dsrupton u u a 2. b a2. b2 2 R a. b a2. b2 R Fgure 6.5 : Example of nequaltes (wea and strong) between fuzzy dual numbers More, a fuzzy equalty wrtten can be defned between two fuzzy dual numbers by: ~ a. b, a. b : a. b a. b a 2 a. b, a. b and a a. b, a. b (6.) u u 2 R 2 R a 2. b2 a. b a 2. b2 a. b Fgure 6.6 : Examples of fuzzy equalty between fuzzy dual numbers Then any two fuzzy dual numbers can be raned as ether strongly dfferent, wealy dfferent or rather equal and a fuzzy ranng can be establshed between them as well as max and mn operators over subsets of ~ Fuzzy dual delays and duratons It s supposed here that t s possble consderng the perturbed stuaton for all future ground handlng related events to propose earlest and latest expected completon tmes, t mn and t max to construct a fuzzy dual trangular completon tme number ~ t where: R ~ t ) ( t mn t ) / 2 and D ~ t ) ( t max t ) / 2 (6.2) ( max ( mn It s also supposed that the duraton of each type of ground handlng tas can be represented n the same way by a fuzzy dual number d ~ : ~ ~ R d ) ( d mn d ) / 2 and D d ) ( d max d ) / 2 (6.3) ( max ( mn 4

162 Chapter 6 Ground Handlng Management Under Dsrupton That means that f at tme t the consdered event requres the avalablty of some equpment or team, an equpment or team of ths type should be planned to be avalable at tme t mn to be sure to avod delay and cannot be reassgned n the plannng wth certanty to any other tas before tme t max. Here dmn and dmax wll be assocated respectvely wth the mnmum and the maxmum dfference between the fnshng and the startng tmes of the correspondng tas. Ths fuzzy dual formalsm s here adopted snce t provdes a smple way to tae nto account operatons uncertanty compared to probablstc approaches and allow straghtforward calculatons and nterpretaton Ranng Flght under Dsrupton wth Uncertanty The followng notatons are adopted: each tas of a ground handlng process,...,ts carred out on an arcraft a() assocated to a flght, I, (I=I A I D, I A s the set of scheduled arrvng flghts durng the next management horzon flghts and I D s the set of scheduled departng flghts durng the same perod) by a specfc ground handlng servce provder,..., K. The frst step of the proposed heurstc conssts n performng an ntal orderng of the flghts scheduled to arrve wthn the next ground handlng management horzon n accordance wth ther current predcted arrval tme a tˆ at ther assgned parng amended by consderng ther crtcalty. To each arrvng flght I A, can be assgned the dfference t tˆ t between a a a the predcted arrval tme tˆ a and the scheduled arrval tme a a t. Here tˆ and t a can be ether real a numbers or fuzzy dual numbers, where tˆ s provded by the ATC. In the second case, ths corresponds practcally to a tme wndow. Each arrvng flght wll cope wth two types of operatonal constrants: - Connecton constrants when arrvng passengers must reach wthout delay others departng flghts. - Departure schedule when the arrvng arcraft must be ready to start a new flght wth a tght schedule. 42

163 Chapter 6 Ground Handlng Management Under Dsrupton Ground handlng at arrval Tax n Unload luggage De-boardng Santaton cleanng Fuellng Caterng Load luggage Boardng passengers Water Ground handlng at departure Pushng bac Tax out When consderng connecton constrants, let Fgure 6.7 : Example of ground handlng actvtes sequencng C be the set of departng flghts connected to arrvng flght. The tme margn between flght and each flght j n C s gven by: a j d j a ~ ddb T ~ j, d ~ ul j j C m~ t tˆ max ~ (6.4) Here T ~ and ~ j j are respectvely the connectng delay for passengers and luggage between flghts and j. The margn between arrval flght and departure flght j servced n mmedate successon by the same arcraft s: m ~ ~ t tˆ D wth j () (6.5) a j d j a j where D ~ s the mnmum fuzzy dual duraton of ground handlng around arrval of flght and j departure of flght j. Here () provdes the number of the next flght servced by the arcraft operatng flght. Then: ~ D j ~ d ~ d ul ~ d ~ d db ca bd max ~ ~ ~ d (6.6) pb d db d cl dbd ~ d sa fu ~ d ~ d ~ d wa ll ~ 43

164 Chapter 6 Ground Handlng Management Under Dsrupton Then, the fuzzy margn of arrvng arcraft s gven by: ~ ~ (6.7) a a m mn mj jc ( ) The amended arrval tme for flght s then gven by: ~ t tˆ m~ (6.8) a a a To each departng flght I D, can be assgned the dfference t tˆ t between the d d d d predcted departure tme tˆ and the scheduled departure tme t d. Here also, d d tˆ and t can be ether real numbers or fuzzy dual numbers. Symmetrcally, each departng flght must cope wth operatonal constrants related wth successve flghts by the same arcraft and flght connectons for passengers and cargo. In the case n whch the ground handlng tass are relatve to a departng flght j, the amended predcted tme to start ground handlng actvtes at the correspondng parng poston s now gven by: ~ t d j t d j mn jc and ( j) m~ a j (6.9) Wth m~ ~ d ~ d fu ll a ~ ~ ~ ( ) max dca dbd d (6.20) pb ~ d wa Then, to each flght, ether arrvng or departng, s assgned a tme parameter such as: a a ~ t for arrvng flghts (6.2.a) d d ~ t for departng flghts (6.2.b) where s the fuzzy dual pseudo norm defned n the appendx. Then the flghts, ether arrvng or departng, present n the consdered perod of operaton can be raned accordng to a ncreasng ndexes and d departng flght.. Let the nteger r a () and r d () be the amended ran of arrvng or 44

165 Chapter 6 Ground Handlng Management Under Dsrupton 6.6. Ground Handlng Fleets assgnment to flghts Then arrvng and departure flghts are processed n the correspondng produced orders r a () and r d (), where ground handlng unts are assgned to the correspondng arcraft. In the case of an arrvng flght, ground handlng arrval tass (unloadng luggage, de-boardng, cleanng and santaton) are coped wth by assgnng the correspondng ground handlng unts n accordance to ther prevous assgned tass wth other arcraft, ther current avalablty, and ther current dstance to the consdered arcraft. Here the common reference tme schedule for a the ground handlng arrval tass s tˆ, I. A In the case of a departng flght, ground handlng departure tass (fuellng, caterng, luggage loadng, boardng, water and push bac) are also coped wth by assgnng the correspondng ground handlng unts n accordance to ther prevous assgned tass wth other arcraft, ther current avalablty, and ther current dstance to the consdered arcraft. Here the common reference tme schedule for the ground handlng departure tass s B low ~ ( t ), I. In both cases t s consdered that the whole set of dfferent ground handlng unts necessary at arrval or departure s assgned by consderng the common reference tme schedule. Ths assgnment of ground handlng unts to flghts ether arrvng or departng s performed on a greedy base by consderng the closest vehcle avalable to perform the requred tas. Ths wll mae that at the start of ground handlng actvtes for an arrval or departure flght, all necessary resources wll be nearby the parng place and that schedulng constrants between elementary ground handlng tass wll be coped wth locally wthout need of communcaton between the dfferent ground handlng managers. Ths s a rather smple greedy heurstc whch provdes for each fleet facng the current servce demand a complete soluton through a reduced computatonal effort. So there s no lmtaton n callng bac ths soluton process any tme a sgnfcant perturbaton occurs. d D 45

166 Chapter 6 Ground Handlng Management Under Dsrupton 6.7 Illustraton of the proposed approach To evaluate the proposed approach, the data used on the study case of the Chapter 5 has been modfed to create artfcally a dsrupton stuaton. Here t has been consdered that for any external reason, for exemple some severe weather condtons, a part of earler scheduled arrvng flghts n the mornng have been delayed and the arport operates under a concentrated arrvng traffc at capacty between a.m. and p.m.. Then, the effectve arrvals and scheduled departures are those of Table 6.. It s consdered that durng and after ths perod the arsde capacty of the arport s nsuffcent, ncludng taxng capacty wth the appearence of queues of taxng arcraft, parng postons wth apron congeston and saturated ground handlng capacty. In that condtons, transfer tmes for arcraft and ground handlng unts actvtes duratons are subject to large uncertantes. Here t has been consdered two scenaros for the uncertanty: n the frst one addtonal delays are between 0% and 40% of the orgnal duraton between a.m. and 2 p.m. wth return to nomnal stuaton afterwards, n the second scenaro addtonal delays are between 0% and 40% of the orgnal duraton between a.m. and noon, between 20% and 60% of the orgnal duraton between noon and :30 p.m., between 0% and 40% of the orgnal duraton between :30 p.m. and 2:30 p.m. wth return to nomnal stuaton afterwards. 0hh h2h 2h3h 3h4h 4h5h 5h6h Arrval traffc Scheduled departures Table6. : Effectve arrvals and scheduled departures In the case of ths arport, there are no connectons between the flghts snce n general ths arport s a fnal destnaton for most of the passengers, so the arrval and the departure prorty lsts concde. The prorty lst s calculated here by tang nto account the predcted departure date of the flght j, whch s the flght servced by the same arcraft than for flght. Here D ~ j s the mnmum fuzzy dual duraton of ground handlng around arrval of flght 46

167 Chapter 6 Ground Handlng Management Under Dsrupton and departure of flght j and the real arrval date of the flght respectng the consderng degree of uncertanty. Ths duraton ~, whch s a fuzzy dual number, can be expressed by: j ~ j ~ D j tˆ a t d j (6.22) Arrval tmes a tˆ Departure tmes d t j D j j Fgure 6.8 : Illustraton of the duraton Ths applcaton provded a feasble assgnment for each ground handlng manager n at most 0.4 seconds each updatng of the prorty lsts. The numercal results show that the delayed arcraft get n general the hghest prorty on the lst. Durng the perod of tme between a.m and 2:30 p.m. ground handlng acheves to serve 200 flghts (arrval and departure of arcraft). The man numercal results are dsplayed n table 6.. Scenaro Scenaro 2 Mean delay for GH processng at arrval 7.36 mn 8.86 mn Maxmum delay for GH processng at arrval 27 mn 30 mn Mean delay for GH processng at departure 45. mn 59.4 mn Maxmum delay for GH processng at departure 95 mn 97 mn ~ Table6. 2:Statstcal results for dsrupton scenaros j Fgure 6.9and 6.0 dsplays the hourly dstrbuton of delayed arcraft at departure resultng from the applcaton of the proposed approach for the two scenaros. It appears that the mpact 47

168 Chapter 6 Ground Handlng Management Under Dsrupton of arrvng traffc delays has resulted n an arport dsrupton stuaton whch has extended n the afternoon. In the frst scenaro t can be consdered that the dsrupton stuaton ends around5 p.m. and n the other case t ends around 9p.m.. It appears then, that the more uncertanty about arsde operatons delays, the less the avalable ground handlng capacty s able to cope wth ths dsrupton stuaton. Then nsurng predctablty of arsde delays through fludty of operatons even n heavy actvty levels stuatons emerge as an mportant objectve. Number of acraft departure delay :00 :00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00 0:00 :00 3:00 4:00 5:00 6:00 7:00 8:00 9:00 20:00 2:00 22:00 23:00 0:00 Tme Fgure 6.9 : The hourly dstrbuton of delayed arcraft at departure (Scenaro ) Number of arcraft departure delay :00 :00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00 0:00 :00 3:00 4:00 5:00 6:00 7:00 8:00 9:00 20:00 2:00 22:00 23:00 0:00 Tme Fgure 6.0 : The hourly dstrbuton of delayed arcraft at departure (Scenaro 2) 48

169 Chapter 6 Ground Handlng Management Under Dsrupton 6.8 Concluson In ths chapter, the proposed framewor for ground handlng management has been consdered n the case of a huge traffc perturbaton characterzng an arport dsrupton. In a frst step the concept of arport dsrupton has been analyzed as well as the man sources of arport dsrupton, and a defnton has been proposed for t. Then the operatons plannng procedures performed wthn the proposed management structure of ground handlng have been revsed by adoptng temporary new objectves and tang nto account the uncertanty wth respect to actvty delays n ths stuaton. Durng the dsrupton perod, the ground handlng coordnator taes over the drecton of the ground handlng management by mposng to the ground handlng managers, prorty lsts of flghts to be processed. The computaton of these prorty lsts maes use of fuzzy dual calculus to tae nto account delays uncertanty. The feasblty of the proposed approach s dsplayed by consderng the case of a dsrupton at Palma de Mallorca arport. 49

170 Chapter 6 Ground Handlng Management Under Dsrupton 50

171 Concluson and Perspectves CONCLUSION AND PERSPECTIVES 5

172 52 Concluson and Perspectves

173 Concluson and Perspectves The sustaned global economc growth of the last decades has been made feasble by the development of mproved means of communcaton and of transportaton of people and goods. It has been partcularly the case wth ar transportaton where, durng the last forty years, the number of passengers has been multpled by seven. Ths ncrease of passenger volume has been possble by a correspondng ncrease of arcraft traffc whch a permanent challenge for cvl avaton authortes and arports to supply suffcent capacty to provde a safe transportaton servce wth acceptable qualty standards. Then, n the last decade, new traffc management practces, such as A-CDM, based on mult-agent and collaboratve decson mang concepts have been ntroduced. Among the many actvtes whch contrbute to the safety and effcency of ar transportaton, arport ground handlng plays an mportant role even f t has not been too much medatsed relatvely to plots and ATC ssues. In ths thess arport ground handlng has been frst descrbed and analyzed, demonstratng the dversty and the complexty of the ground handlng actvtes performed on a grounded arcraft whch are organzed n a seral-parallel structure where any delay on a partcular actvty may have a strong mpact on ts overall performance. It has appeared that to avod delays generated by ground handlng actvtes, there s a need for a tght synchronzaton to process the stream of arrvng/departng arcraft. Then ths ntroduces the need for an effcent management structure to mantan ths whole process n effcency grounds and contrbute postvely to the arport performance. Consdered the actual practce t has been found that the concerned staeholders (arport authortes, arlnes, specalzed ground handlng operators) are today nvolved n varable degrees n the management of ground handlng at dfferent large arports. Also, t has been observed that when consderng drect and ndrect costs related to ground handlng at arports, drect cost resultng from the executon of ground handlng tass are relatvely very small wth respect to potental over costs resultng from even lmted dysfunctons of ground handlng operatons. Then t has appeared crucal to promote the ablty of the ground handlng management to be able to prevent dsruptons and to reduce the mpact of traffc perturbatons when they happen. Ths supposes the avalablty of the rght decson processed wthn the rght management organzaton. An overvew of the man decson processes developed n the feld of Operatons Research, n general formulated as optmzaton problems, has been performed, showng the dffculty to adopt n that case exact soluton approaches ether 53

174 Concluson and Perspectves for the management of a partcular ground handlng actvty or for an overall optmzaton of the ground handlng resource assgnment and schedulng. Next, dfferent heurstcs have been bult to provde a soluton to these nomnal problems, however few wors had been done n report some experments where the heurstc appled to ground handlng schedulng are assessed n perturbed envronments. Ether usng exact or approxmate methods, t appeared that the many of these studes mss to consder the cost dmenson where the drect cost resultng from ground handlng actvtes s secondary wth respect to the economc consequences of delays at servcng arrvng and departng arcraft and the management dmenson where an organzaton able to cope wth routne stuatons as well as perturbed condtons or even dsrupted stuatons, must be desgned. Then, t has been shown that adoptng a herarchcal approach, t s possble to organze ground handlng management n accordance wth the A-CDM approach where a ground handlng coordnator operates as an actve nterface between the ar transport operators and the specalzed ground handlng managers n charge of the ground handlng unts. The nformaton flows assocated wth the dfferent levels of management and operatons have been descrbed usng the Petr net formalsm. Then, the dfferent assgnment problems solved by the ground handlng coordnator and ground handlng managers have been consdered. Consderng the complexty of the respectve problems, greedy heurstcs have been chosen to llustrate the proposed approach. The whole process has been llustrated frst by consderng a case study wth real traffc presentng rather lmted perturbatons. Then n a second step, the proposed framewor for ground handlng management has been consdered n the case of a huge traffc perturbaton characterzng an arport dsrupton. The man objectve of ths PhD thess has been to contrbute to the desgn of a general effcent management organzaton for ground handlng at arports. Many perspectves of research and development amed at mprovng the arport performance when consderng the ground handlng sector, arse n dfferent felds to complete the present study: The collaboratve decson mang process used by the A-CDM partners should ntegrate the proposed organzaton of the ground handlng management functon wth the ground handlng coordnator as nterface. The capablty of the ground handlng coordnator to perform hs tass should be based on mproved decson processes coverng ssues such as: 54

175 Concluson and Perspectves - Evaluaton of the mpact of operatonal perturbatons and the effectveness of ground handlng to cope wth them, - Processng of nformaton wth a varable degree of uncertanty, - Montorng and dagnostc of the overall ground handlng process by detectng abnormal operatonal stuatons up to dsrupton, - Adaptng operatonal objectves accordng to stuaton dagnostc and prorty ranng of flght to be processed by ground handlng, - Dynamc szng of reserve ground handlng resources, - Generaton of overall bac-up solutons for ground handlng resource assgnment, - Predcton of mlestones to be communcated to the other A-CDM partners. The capablty of each ground handlng manager to assgn effcently, accordng to the drectves of the ground handlng coordnator, ether at the parng level or the roaster level, hs assgnment of avalable resources to the dfferent ground handlng tass should be based on mproved decson processes. The present study has made some general assumptons about the arport and the traffc consdered, whle each arport has ts own characterstcs. Thus any general framewor to manage the whole or a part of ground handlng management e, should be partcularzed at the development level. 55

176 56 Concluson and Perspectves

177 Bblographes BIBLIOGRAPHY 57

178 58 Bblographes

179 Bblographes [Arbus, 2005]: Arbus, AXXX arplane characterstcs for arport plannng, the Arbus Company, 2005 [Ashford and al. 203]: N.J. Ashford, H.P. Martn Staton, C.A. Moore, P. Coutu, J.R. Beasley, Arport Operatons, Thrd edton, McGraw Hll, pp:53-80, 203. [Bllonnet, 2007]: A. Bllonnet, Optmsaton dscrète: de la modélsaton à la résoluton par des logcels de programmaton mathématque, pp: , Dunod, Pars, 2007 [Boeng, 2009]: Boeng, BXXX arplane characterstcs for arport plannng, the Boeng Company, 2009 [Bratu and al, 2006]: S. Bratu and S. Barnha, Flght operatons recovery: New approaches consderng passenger recovery, 2006, Journal of Schedulng, Vol.9, No. 3, pp [Brucer, 999] : P. Brucer, A. Drexl, R. Mohrng, K. Neumann, Resource-constraned project schedulng: Notaton, classfcaton, models, and methods, European Journal of Operatonal Research 2, pp: 3-4, 999. [Busness contnuty management, 2006]:Clause 2.3 BS Busness contnuty management, Brtsh Standards Insttute [Clausen and al, 2005]: J. Clausen, A. Larsen, J. Larsen, Dsrupton management n the arlne ndustry Concepts, models and methods, Techncal report, 2005, Informatcs and Mathematcal Modellng, Techncal Unversty of Denmar, DTU. [Clausen, 20]: T. Clausen, Arport Ground Staff Schedulng, Ph.D Thess, DTU Management Engneerng, Techncal Unversty of Denmar, March 20. [Cosenza and al, 20]: C.A.N. Cosenza, F. Mora-Camno, Nombres et ensembles duaux flous et applcatons, Techncal Repport, LMF Labortory, COPPE/UFRJ, Ro de Janero, August 20. [Cosenza and al, 202]: C.A.N. Cosenza, O. Lenguere, F. Mora-Camno, Fuzzy sets and dual numbers: an ntegrated approach, Proceedngs of 9 th Internatonal Conference on Fuzzy Sets and Knowledge Dscovery, Chonqng, pp: 8-86, 202. [Desaulners and al, 2005]: G. Desaulners, J. Desrosers and M. Solomon, Column Generaton Sprng-Verlag, New Yor n 9th of Ma 2005 p385 [Dohn and al, 2008]: A. Dohn, E. Kolnd, Optmzng manpower allocaton for ground handlng tass n arports usng column generaton, Proceedngs 43 rd Annual Conference of the Operaton Research Socety of New Zealand, pp: 2-,

180 Bblographes [Dorgo and al, 997]: M. Dorgo, L. M. Gambardella, Ant Colony System: A Cooperatve Learnng Approach to the Travelng Salesman Problem, IEEE Trans. Evolutonary Computaton, Vol., No., pp , Aprl, 997. [Du and al, 2008]: Y. Du, Q. Zhang, Q. Chen, ACO-IH: An Improved Ant Colony Optmzaton Algorthm for Arport Ground Servce Schedulng, IEEE Internatonal Conference on Industral Technology, pp:-6, 2008 [Dyer and al, 2003]: M. Dyer, L. Stouge, Computaonal complexty of stochastc programmng problems, SPOR-Report , Dept. of Mathematcs and Computer Scence, Endhoven Techncal Unversty, Edhoven, [Eurocontorol, 203]: www. euro-cdm.org [Feber, 995] : J. Feber, Les systèmes mult-agents : Vers une ntellgence collectve, InterEdtons, 995. [Feo and al, 995] Feo T A and M G C Resende, 995 Greedy Randomzed Adaptve Search Procedures, Journal of Global Optmzaton, 6, pp [Frce and al, 2009]: H. Frce and M. Schultz. Delay mpacts onto turnaround performance Optmal tme bufferng for mnmzng delay propagaton. Eghth USA/Europe Ar Traffc Management Research and Development Semnar (ATM), 2009 [G. Nagy and al, 2005] :G. Nagy and S. Sald. Heurstc algorthms for sngle and multple depot vehcle routng problems wth pcups and delveres. European Journal of Operatonal Research, 62:26 4, [Garca and al, 20]: P. Garca Ansola, A. Garca Hguera, J.M. Pastor, F.J. Otamend, Agent-based decson-mang process n arport ground handlng management, Logst. Res. 3, pp: 33-43, 20. [Ghana and al, 2000]: G. Ghana and G. Improta. An effcent transformaton of the generalzed vehcle routng problem.european Journal of Operatons Research, 22: 7, [Gwggner, 2004]: C. Gwggner, Implct relaton between tme slots, capacty and real demand n ATFM, the 23 rd Dgtal Avoncs System Conference (Volume:), pp: 3.C Vol., 2004 [Ho and al, 2009]: C. Ho, J.M.Y. Leung, Solvng a manpower schedulng problem for arlne caterng usng metaheurstcs, European Journal of Operatonal Research 202,pp: ,

181 Bblographes [Janc, 997]: M. Janc, The flow management problem n ar traffc control: a model of assgnng prortes for landngs at a congested arport, Transportaton Plannng and Technology 20, pp: 3-62, 997. [Kohl and al, 2007]: N. Kohl, A. Laresen, J. Larsen, A. Ross, S. Tourne, Arlne dsrupton management- Perspectves, experences and outloo, 2007, Journal of Ar Transport Management 3, [Kosnsy, 2006]: Kosnsy W., (2006), On Fuzzy Number Calculus, Internatonal Journal of Appled Mathematcs and Computer Scence, Vol.6, No., [Kuhn and al 2009]:K. Kuhn and S. Loth, Arport Servce Vehcle Schedulng, Eghth USA/ Europe Ar Traffc Management Research and Development Semnar (ATM), [Land and al, 960] : A.H. Land and A.G. Dog (960) An automatc method of solvng dscrete programmng problems. Econometrva 28 (3).pp [Lettovsy, 997]: L. Lettocsy, Arlne operatons recovery: An optmzaton approach, 997, Ph.D dssertaton, Georga Insttute of Technology, Atlanta, USA. [Mao and al, 2008]: X.Mao, A.Mors, N.Roos, C.Wtteveen, Agent- based schedulng for arcraft de-cng, [Mao and al, 2009]: X. Mao, N. Roos, A. Salden, Stable Mult-project Schedulng of Arport Ground Handlng Servces by Heterogeneous Agents, 8th Internatonal Conference on Autonomous Agents and Multagent System, 0-5 May, 2009, Budapest, Hungary. [Nasser, 2006]: Nasser H, (2006), Fuzzy Number: Postve and Nonnegatve, Internatonal Mathematcal Forum, vol.3, [Norn and al, 2009]: A.Norn, T. Andersson Granberg, P. Vabrand, D. Yuan, Integratng optmzaton and smulaton to gan more effcent arport logstcs, Eghth USA/ Europe Ar Traffc Management Research and Development Semnar (ATM), [PDM, 202]: "The Aena, Palma de Mallorca Arport", [Psnger and al, 2007]: D. Psnger and S. Rope. A general heurstc for vehcle routng problems. Computers & Operatons Research, 34(8): , 2007], 6

182 Bblographes [Ploog, b2005]: D. Ploog, Dsrupton Management n Operaton Control, m2p Consultng, Presentaton, Manz 2005 [Prns, 2002]: C. Prns, Effcent heurstcs for the heterogeous fleet multtrp vrp wth applcaton to a large-scale real case. Journal of Mathematcal Modellng and Algorthms, :35 50, 2002.] [Rav and al, 2004]: R. Rav, A. Snha, Hedgng uncertanty: approxmaton algorthms for stochastc optmzaton problems, In Proceedngs of 0 th IPCO, pages 0-5, [Rchetta and al 993]: O. Rchetta, A.R. Odon, Solvng optmally the statc ground holdng polcy problem n ar traffc control, Transportaton Scence 27, pp: , 993. [Ronchetto, 2006]: C. Ronchetto, The costs of delays and cancellatons Analyss and means for cost reductons, m2p Consultng, Presentaton, AGIFORS Duba [Rope and al, 2006]:S. Rope and D. Psnger. A unfed heurstc for a large class of vehcle routng problems wth bachauls. European Journal of Operatonal Research, 7: , [Tanger and al, 203]: R. Tanger and E. Clayton, Booz & company s London and Kuala Lumpur offces. Managng Arport Dsrupton: Achevng Reslence through Collaboraton (203). [Toth and al, 2002] P. Toth and D. Vgo. An overvew of vehcle routng problems. In P. Toth and D. Vgo, edtors, TheVehcle Routng Problem, volume 9 of SIAMMonographs on Dscrete Mathematcs and Applcatons, chapter, pages 26. SIAM, Phladelpha, [Vdosavljevć and al, 200]: A. Vdosavljevc, V. Tosc, Modelng of turnaround process usng Petr Nets, Dvson of Arports and Ar Traffc Safety (APATC), Unversty of Belgrade, Faculty of Transport and Traffc Engneerng (FTTE), (0/0/203) [Wu and al 2000]: C.Wu, R.E. Caves, Arcraft operatonal costs and turnaround effcency at arports, Journal of Ar Transport Management 6, pp: , [Zhwe and al, 200]: X. Zhwe, L. Y, Research of algorthms for arcraft ground decng operaton schedulng model, Proceedng of the 8 th word congress on Intellgent Control and automaton, July , Jnan, Chna. [Zurch Arport]: (0/02/203) 62

183 ANNEX I Typcal Tmes For Ground Handlng Actvtes at Ramp ANNEX I TYPICAL TIMES FOR GROUND HANDLING ACTIVITIES AT RAMP (Arplane characterstcs for arport plannng: Arbus Company 2005, BoengCompany 2009) 63

184 ANNEX I Typcal Tmes For Ground Handlng Actvtes at Ramp 64

185 ANNEX I Typcal Tmes For Ground Handlng Actvtes at Ramp Ths annex provdes the typcal tmes for ramp actvtes durng arcraft turn round for dfferent transportaton arcraft. Actual tmes may vary due to specfc practces and operatng condtons. I. A A Full Servcng Turnaround Charts Assumptons for 48 mnutes turnaround chart for full Servcng. Ths turnaround tme s an assumpton regardng a gven example. a. Passenger handlng: Number of passenger: 50 pax, Number of used brdge: brdge () De-boardng: L: 50, 2L:0, - De-boardng rate: 22 pax / mn per door. (2) Boardng: L: 50, 2L:0, - Boardng rate: 8 pax / mn per door. b. Caterng: R - R 2 / sequental, Galley M: 4 FSTE, Galley M2: 7 FSTE c. Cleanng: Tme avalable d. Refuel: 5.6 tons, 734 (l), 2 hoses ( sde) e. Water servcng: 00% f. Tolet servcng: 00% g. Other ground handlng operatons: Securty/Safety checs: Yes (4 mn each) Cabn crew change: Yes (4 mn) Cargo: 2 Cargo loaders, Belt loader, operator / BL, No sldng carpet, FWD compartment: 3 LD3, AFT compartment: 4 LD3, Bul n bul CC:000 g 65

186 ANNEX I Typcal Tmes For Ground Handlng Actvtes at Ramp Fgure I. : Full Servcng Turnaround Charts for an A320-00/ A Mnmum Servcng Turnaround Chart Assumptons for 23 mnutes turnaround chart for a mnmum servcng. Ths turnaround tme s an assumpton regardng a gven example. a. Passenger handlng: 80 pax / 2 starways () De-boardng: L:90, 2L:90, De-boardng rate: 20 pax / mn per door. (2) Boardng: L:90, 2L:90, - Boardng rate: 5 pax / mn per door. b. Caterng: No c. Cleanng: No d. Refuel: 5.6 tons, 734 (l), 2 hoses ( sde) e. Water servcng: 0%: f. Tolet servcng: 0% 66

187 ANNEX I Typcal Tmes For Ground Handlng Actvtes at Ramp g. Other ground handlng operatons: Securty/Safety checs: Yes (4 mn each) Cabn crew change: No Cargo: 2 Cargo loaders, Belt loader, operator / BL, No sldng carpet, FWD compartment bul: 3 LD3, AFT compartment bul: 4 LD3, Bul n bul CC: 00, 00 Fgure I. 2: Mnmum Servcng Turnaround Charts for an A320-00/ A II. A Full Servcng Turn Round Charts Assumptons for full servcng turn round chart. a. Passenger Boardng/De-boardng :De-boardng : 23 passengers (0 frst + 42 busness + 79 toursts), For full servcng, all passengers de-board and board, Doors used: L + L2. () De-boardng: 04 pax at L (0 frst + 42 busness + 52 toursts) and 27 pax at L2, De-boardng rate = 25 pax/mn, Prorty de-boardng for premum passengers 67

188 ANNEX I Typcal Tmes For Ground Handlng Actvtes at Ramp (2)Boardng: 52 pax at L and 79 pax at L2, Boardng rate = 5 pax/mn, Last Pax Seatng Allowance (LPS) + head countng = + 4 mn b. Fuellng: Bloc fuel for Nomnal Range through 4 nozzles, l ( US gal) at 50 ps, Dspenser postonng or removal = 3 mn (fuel truc change) / f any = 5 mn. c. Cleanng: - Cleanng s performed n avalable tme d. Caterng: -3 caterng vehcles, - 36 Full sze trolley: 7 FST at R, 9 FST at R2 and 20 FST at R4, FST exchange tme =.5 mn/fst e. Potable water servcng: Replensh 700 l (85 US gal); flow rate: 60 l/mn (5.85 USgal/mn) f. Waste water servcng (dranng + rnsng): Dscharge 700 l (85 US gal) g. Other ground handlng operatons: Cargo: 6 LD3 + 2 pallets for AFT CC, 8 LD3 + 2 pallets for FWD CC, 000 g (2 205 lb) n Bul CC, LD-3 off-loadng/loadng tmes: off-loadng =.2 mn/ld-3, loadng =.4 mn/ld-3. Pallet loadng tmes: off-loadng = 2.4 mn/pallet, loadng = 2.8 mn/pallet -Bul off-loadng/loadng tmes: off-loadng = 9.2 mn/t, loadng = 0.5 mn/t Start of operatons :() Brdges = t0 = 0, (2) Others = t0 + mn Vehcle postonng/removal = 2 mn (fuel truc excluded) Ground Power Unt (GPU) = up to 2 90 VA, - Ar condtonng = two carts, Dolles per tractor = 4 68

189 ANNEX I Typcal Tmes For Ground Handlng Actvtes at Ramp Fgure I. 3: Full Servcng Turnaround Charts for an A Mnmum Servcng Turnaround Chart Assumptons for 39 mnutes of transt turnaround chart. a. Passenger Boardng/De-boardng : De-boardng : 23 passengers (0 frst + 42 busness + 79 toursts), 50% pax n transt, all passengers de-board and board, Doors used: L + L2 () De-boardng: 04 pax at L (0 frst + 42 busness and 52 toursts) and 27 pax at L2, De-boardng rate = 25 pax/mn, Prorty de-boardng for premum passengers (2) Boardng: 52 pax at L and 79 pax at L2, Boardng rate = 5 pax/mn, Last Pax Seatng Allowance (LPS) + headcountng = + 4 mn b. Fuellng: Refuelng through 2 nozzles, For transt, fuel uplft s 30% of maxmum fuel uplft. (Max = l ( US gal)), Note: local rules and regulatons to be respected, Passengers boardng can start before refuel s fnshed, Dspenser postonng or removal = 3 mn (fuel truc change) / f any = 5 mn 69

190 ANNEX I Typcal Tmes For Ground Handlng Actvtes at Ramp c. Cleanng: Cleanng s performed n avalable tme d. Caterng: Tme needed just for addtonal meals, Assumptons: 0 mn e. Potable water servcng: No f. Waste water servcng: No g. Other ground handlng operatons: Cargo: For transt, 50% of luggage are exchanged n one cargo compartment only, contaner loader for AFT CC, 4 LD3 for AFT CC. LD-3 off-loadng/loadng tmes: off-loadng =.2 mn/ld-3, loadng =.4 mn/ld-3 Start of operatons: Brdges = t0 = 0, Others = t0 + mn, Vehcle postonng/removal = 2 mn (fuel truc excluded), Ground Power Unt (GPU) = up to 2 90 VA, Ar condtonng = two carts, Dolles per tractor = 4 Fgure I. 4: Mnmum Servcng Turnaround Charts for an A

191 ANNEX I Typcal Tmes For Ground Handlng Actvtes at Ramp III. A Models a. Passenger Boardng/De-boardng : 00% (555 pax) passenger exchange: - Doors (type A - 42 wde) used: ML and M2L (man dec) and UR (upper dec). - PB/D rate: boardng = 5 pax/mn / de-boardng = 25 pax/mn - Last Pax Seatng Allowance (LPS) = + 4 mn - 60 star flow rate: up-flow = 4 pax/mn / down-flow = 8 pax/mn b. Fuellng: Bloc fuel for Nomnal Range through 4 nozzles: lters ( US gallons) at 40 ps (48 mn), Dspenser postonng or removal = 3 mn (fuel truc change) / f any = 5 mn c. Cleanng:Full cleanng d. Caterng:Crew adapted to match caterng tme, Full caterng: Average truc capacty = 30 Full Sze Trolley Equvalent (FSTE), Smultaneous caterng and PB/D = not represented, Inbound/outbound FSTE = mxed n the same truc, FSTE exchange tme: Dedcated door-galley =.5 mn/fste, cart crculaton ( Seat zone) = mn/fste, cart crculaton (> Seat zone) = +.0 mn/fste, Va lft: Dedcated door to sngle lft = 2.0 mn/fste e. Potable water (standard/opton) : 700/2 500 ltters (495/660 US gal) at 60 l/mn(23 US gal/mn). f. Waste water: Dscharge and rnsng g. Other ground handlng operatons: Cargo: Full LD-3 exchange (22 + 6) LD-3 and bul exchange of g (4 409 lb) :LD-3 offloadng/loadng tmes: off-loadng =.4 mn/ld-3 / loadng =.7 mn/ld-3, Pallet loadng tmes: off-loadng = 2.5 mn/pallet / loadng = 2.9 mn/pallet, Bul off-loadng/loadng tmes : off-loadng = 9.2 mn/t / loadng = 0.5 mn/t Start of operatons: Brdges = t0 = 0, Others = t0 + mn Vehcle postonng/removal = 2 mn (fuel truc excluded), Upper dec vehcle postonng/removal = 3 mn Clearance between GSE = 0.5 m (20 n) Ground Power Unt (GPU) = up to 4 90 VA, Ar condtonng = two carts, Dolles per tractor = 4 to 6 7

192 ANNEX I Typcal Tmes For Ground Handlng Actvtes at Ramp Fgure I. 5: Mnmum Servcng Turnaround Charts for an A Fgure I. 6: Full Servcng Turnaround Charts for an A

193 ANNEX I Typcal Tmes For Ground Handlng Actvtes at Ramp IV. B LR Models Fgure I. 7: Full Servcng Turnaround Charts for a B LR Fgure I. 8: Mnmum Servcng Turnaround Charts for a B LR 73

194 ANNEX I Typcal Tmes For Ground Handlng Actvtes at Ramp V. B Models Fgure I. 9: Full Servcng Turnaround Charts for a B Fgure I. 0: Mnmum Servcng Turnaround Charts for a B

195 ANNEX II Ground Handlng Fees n Talln Arport ANNEX II GROUND HANDLING FEES IN TALLIN AIRPORT ( 75

196 ANNEX II Ground Handlng Fees n Talln Arport 76

197 ANNEX II Ground Handlng Fees n Talln Arport Ground handlng fees n Tallnn arport Effectve from st of October 202 Servce Unt EUR Basc ground handlng servce Up to 0 MTOW ton 0.00 Over 0-20 MTOW ton 8.50 Over MTOW ton 6.50 Over MTOW ton 5.50 Over MTOW ton 4.00 Over 00 MTOW ton 3.50 Weght & Balance calculaton Per turnaround Passenger and baggage servce Each departng pax Each arrvng pax Man power Hour/ Call Meetng and postonng the arcraft Each MTOW ton 0.60 Arcraft nteror cleanng (up to 50 seats) Call Arcraft nteror cleanng (up to 00 seats) Call Arcraft nteror cleanng (up to 50 seats) Call Arcraft nteror cleanng (up to 200 seats) Call Arcraft nteror cleanng (up to 300 seats) Call Arcraft nteror cleanng (up to 360 seats) Call Arcraft nteror cleanng (up to 440 seats) Call 0.00 Arcraft nteror cleanng (up to 500 seats) Call Arcraft nteror nght-stop cleanng (up to Call seats) Arcraft nteror nght-stop cleanng (up to Call seats) Arcraft nteror nght-stop cleanng (up to Call seats) Arcraft nteror nght-stop cleanng (up to Call seats) Arcraft nteror nght-stop cleanng (up to Call seats) Arcraft nteror nght-stop cleanng (up to Call seats) Arcraft nteror nght-stop cleanng (up to Call seats) Arcraft nteror nght-stop cleanng (up to Call seats) Ltter dspose Each MTOW ton 0.40 Power supply (220V) Hour / Call 5.00 Ground Power Unt Hour / Call Moble Ground Power Unt 28V/5V Hour / Call

198 ANNEX II Ground Handlng Fees n Talln Arport Push bac Call Passenger stars/arbrdge Hour / Call Tolet servce (empty and fll) Call Tolet servce (empty) Call Tolet servce (fll) Call Tolet servce each tan Per tan Water supply Call Dranng water tans Call Heater Hour / Call ASU Hour / Call Addtonal platform for CRJ Call Hghloader-transporter (mx lftng weght Hour / Call t/heght 3.6m Cargo Hghloader (mx lftng weght Hour / Call t/heght 5.6m) Escort on the ramp Hour / Call Crew transport on the ramp Hour / Call Crew cty transport (up to 8 seats) Hour / Call Crew cty transport (over to 8 seats) Hour / Call Hotel boong boong 5.00 Cargo landng g 0.07 Porter servce n passenger termnal -6 pax Each addtonal pax 2.50 Group over 30 each pax 3.00 Equpment rent Towbar Hour / Call Forlft (maw weght 7.5t) Hour / Call Forlft slave pallet Hour / Call LD, LD2, LD3 contaner dolly Hour / Call cargo pallet dolly Hour / Call 5.00 Baggage tractor Hour / Call Baggage cart Hour / Call 0.00 Belt-loader Hour / Call Hanger rent De-/ Ant-Icg Call / Group A-B Wng span up to 23m Call / Group C Wng span 24-35m Call / Group D Wng span 36-5m Call / Group E Wng span 52-65m 78 To be agreed (depend on MTOW) Type (mxture) lter 3.40 Type 2 lter 4.40 Manual snow removal Call Arport and navgaton fees Landng fee Each MTOW ton 8.3

199 ANNEX II Ground Handlng Fees n Talln Arport Passenger fee Each departng pax 7.03 Parng fee: Free parng up to 6 hours all cargo arcraft Free parng up to 3 hours all other arcraft Each MTOW ton per 24 hours for nonbased arcraft.53 79

200 ANNEX II Ground Handlng Fees n Talln Arport 80

201 ANNEX III The Vehcle Routng Problem Wth Tme Wndow ANNEX III THE VEHICLE ROUTING PROBLEM WITH TIME WINDOWS (Lenstra & Kan, 98) 8

202 ANNEX III The Vehcle Routng Problem Wth Tme Wndow 82

203 ANNEX III The Vehcle Routng Problem Wth Tme Wndow. Theoretcal bacground The Vehcle Routng Problem (VRP) s one of the most popular combnatoral optmzaton problems. It s amed at determnng an optmal set of routes for an avalable fleet of vehcles n order to servce a set of customers, subject to dfferent constrants. Wth many other related problems, t s NP-Hard (Lenstra & Kan, 98)and besde exact methods, many heurstcs approaches have been developed.. The Vehcle Routng Problem wth Tme Wndows (VRPTW) and Vehcle Routng and Schedulng Problem (VRSP) are extensons of the VRP to turn t more realstc. In the case of the VRPTW each customer has a tme wndow wthn whch the vehcle has to begn the servce and for the VRSP there are precedence constrants between the costumers. 2. Mathematcal formulaton: The VRPTW can be formally stated as follows: gven the graph G ( V, A), where V 0,...,n denotes the set of all vertces n the graph representng the ctes wth the depot located at the vertex 0, K s the set of avalable vehcles that can be used and A s the set of arcs. Each arc, j j s assocated a non-negatve dstance matrx C c j whch can be nterpreted as a travel cost or as a travel tme. Gven K the set of avalable vehcles to be routed and scheduled. A nonnegatve demand d, a servce tme servce should be start are assocated to each costumer l s and a tme wndow e, C n whch the. e s the earlest servce tme and l s the latest servce tme allowed to serve the costumer. Each arc has a cost C and a travel j tme t j. At each costumer, the servce start tme must be wthn the tme wndow. Each vehcle must leave and return to the depot after servcng all ts customers. served the costumer j after servng the costumer, and 0 otherwse. x, j Equal f vehcle b s the start tme at whch the vehcle begn to serve the costumer. so the VRPTW conssts to fnd a route wth a least cost and respectng the followng constrants:. Each costumer s served exactly once by exactng one vehcle respectng the tme wndow. 2. All vehcle routes start and end at the depot. K N j j N x j N () x (2) 0 j K 83

204 ANNEX III The Vehcle Routng Problem Wth Tme Wndow N e j N b x (3) x 0 K h l jn x hj h N K (4) N (5) j xj b b s t. N, j N, K (6) Constrant () states that each costumer has to be vsted exactly once, the constrant (2) and (3) state that the servce of each vehcle starts and ends at the depot, the constrant (4) s a flow-balance constrant; f a vehcle arrves at a costumer, t must leave that costumer next. The wndow tme s showed n the constrant (5) and the constrant (6) descrbed the fact that the vehcle cannot start servng a costumer snce t has not fnshed servcng the precedent one. The VRPTW as t has been sad before t s a generalzaton of the VRP. It can be consdered also as a combnaton between the VRP and the schedulng problem or as t nown as the Vehcle Routng and Schedulng Problem whch tae place n many real world applcatons. 3. Soluton approaches The VRPTW has been extensvely studed and several formulatons, exact algorthms, heurstcs and metaheurstcs have been proposed n the past decades. 3. Exact methods for the VRPTW The exact approaches can be classfed to: Lagrange Decomposton based methods: Varous Lagrangan decomposton schemes have been appled to the VRPTW n order to fnd lower bounds. Jornsten and al (986), Madsen (988, 990) and Hales (992) were the most nterest wors whch treat ths subject wth ths approach. Accordng Marshall and al (995) they can currently fnd the optmal soluton of 00 customer problems usng a combnaton of Lagrangan decomposton and branch- and bound. K-tree based methods Fsher and al (997), Holland (975) and Kolh and al (997) used the -tree approach followed by Lagrangan relaxaton to solve ths problem. Fsher and al (997) proposed an algorthm to solve the VRPTW optmally by formulatng the problem as a K-tree problem 84

205 ANNEX III The Vehcle Routng Problem Wth Tme Wndow wth degree 2K on the depot. They consdered that a K-tree for a graph contanng n+ vertces could be presented as a set of n+k edges spannng the graph. So, the problem was solved as fndng a K-tree wth degree 2K on the depot, degree 2 on the customers and subject to tme and capacty constrants. A K-tree wth degree 2K on the depot n ths context s proportonal to K routes. Approaches based on Column Generaton Desrrosers and al (984) s the frst study that has used the column generaton to solve the VRPTW. They amelorated t and n 992 they presented an exact method able to solve 00- costumers problems. Ths method s a combnaton of lnear programmng relaxed set coverng and column generaton. Approaches based on Dynamc Programmng The dynamc programmng approach has been used to solve the VRPTW for the frst tme by Kolen and al (987), and they were based, n ther study, on the wor of Chrstofdes and al (984) who used the dynamc programmng approach to solve the VRP. The problems up to 5 customers are solved to optmalty. 3.2 Heurstc algorthms Route-buldng heurstcs [Baer and al, 989] was the frst paper that proposed a route-buldng heurstcs for the VRPTW. The proposed algorthm conssts, frstly, to defne all possble sngle-costumer routes, and secondly, to determne for each teratonthe two routes whose combnaton provdes the maxmum savng. There the savng s defned as the sum of the tme at whch the vehcle quts the customer to arrve at the depot and the tme at whch the vehcle quts the depot to arrves to customer j and the route form factor. On the bass ths algorthm [Baer and al, 980]elaborated a tme orented nearest- neghbourhood algorthm. The consdered savng was defned as a combnaton of dstance, tme and tme untl feasblty. Another approach presented n [Antes and al, 995] bult upon the nserton dea where each unserved customer ased to be served. Each vehcles n the schedule and whch receved from these unserved costumers a savng for nserton. Then these customers propose to the vehcles ther best offer whch wll be accepted by the vehcles f are the best accordng ther routes consderng the number of alternatves. The customers van be removed from the vehcles routes f they volate the threshold of vehcles routes s volated a certan number of customers are removed and the process s ntated agan. 85

206 ANNEX III The Vehcle Routng Problem Wth Tme Wndow Neghborhood based heurstcs The r-opt s a heurstc whch s based on the neghbors to solve the routng and schedulng problems. Ths heurstc conssts n removng r arcs from ther current soluton and replacng them by other r arcs. The optmal soluton r-optmal s obtaned when the r-opt neghborhood have been used and t cannot be mproved more. In general, r s at most 3, but t s has been proved that t was dffcult to use ths number to solve the VRPTW problem snce t leads to a volaton of the tme wndows [Potvn and al, 995]. [Potvn and al, 995], to solve the VRPTW, used the 2-Opt*. [Chrstofdes and al, 984] solved the VRP usng the - node nterchange. Ths wor has been a reference to solve the VRPTW by many others researchers. The λ- nterchange has been proposed by [Osman, 993] to solve the VRP whch consdered as a base to solve the VRPTW by other authors [ref]. Fnally, [Schulze and al, 999] adopted the shft-sequence neghborhood operator to fnd a soluton for the VRPTW. 3.3 Metaheurstcs Smulated annealng In [Chang and al, 996] three dfferent smulated annealng have been consdered to solve the VRPTW: the frst usng the -node nterchange neghborhood operator, the second usng the λ- nterchange neghborhood operator presented n [Osman, 993] and the thrd usng an algorthm whch adopted the concept of the tabu lst (tabu search metaheurstc). The results showed that the second and the thrd converged faster than the frst one. The three of these methods gave a soluton n whch the dstances travelled were between 7% and % from the optmum. [Thamgah and al, 995] used a non-monotone probablty functon and the λ- nterchange neghborhood operator wth decreasng the temperature n each teraton. The solutons obtaned n ths wor had the same qualty as those obtaned n [Chang and al, 996]. Tabu search The parallelzaton of the tabu search has been used to solve the VRPTW by many authors. In [Garca and al, 994], to fnd the frst soluton, the authors used the Solomon heurstc and the 2-opt* and Or-opt as neghborhood operators. Here, the neghborhood was restrcted to arcs close n dstance. [Badeau a,d al, 995] used the same heurstc to fnd the ntal soluton but combned wth the cross neghborhood operator. [Cordeau and al, 200] adopted the modfcaton of Sweep heurstc to fnd the ntal soluton and the relocate and GENI as neghborhood operator, n ths wor the nfeasbltes were allowed durng the 86

207 ANNEX III The Vehcle Routng Problem Wth Tme Wndow search. [Gehrng and al, 200] solve the problem by consderng the savngs heurstc for the ntal soluton, the Or-opt, 2-opt* and the λ- nterchange as neghborhood operators, the tabu search had been hybrdzed wth an evolutonary algorthm. Generally, and accordng to many wors, the tabu search have been consdered as best heurstcs for the VRPTW. One of the conclusons n [Badeau and al, 995] s that dversfcaton/ ntensfcaton s just as mportant n obtanng good solutons as varable length tabu lst. Genetc algorthm [Thagah and al, 99] was the frst paper usng the genetc algorthm to solve the VRPTW. In ths wor, the genetc algorthm was adopted to fnd good clusters of customers, accordng to a cluster-frst and a route-second problem-solvng strategy. Snce the appearance of the frst paper, many other wors have been adopted ths metaheurstc to solve the VRPTW and whch provded good solutons. Generally, the most of these wors used a hybrd presentaton of the genetc algorthm by consderng: - dfferent heurstc constructon as [Blanton and al, 993], [Berger and al, 998]), - local search ([Thangah, 995a, b], [Thangah and al, 995], [Potvn and al, 996]; [Jung and al, 2002]), - tabu search [Kt and al, 200] - ant colony systems [Berger and al, 2003]. [Antes and al, 995] :Antes, J., Dergs, U., 995. A New Parallel Tour Constructon Algorthm for the Vehcle Routng Problem wth Tme Wndows. Techncal Report, Lehrstuhl fur Wrtschaftsnformat und Operatons Research, Unverstat zu Koln. [Badeau and al, 995] :Badeau, P., Gendreau, M., Guertn, F., Potvn, J., Tallard, E., 995. A Parallel Tabu Search Heurstc for the Vehcle Routng Problem wth Tme Wndows.Techncal Report, CRT-95-84, Centre de recherche sur les transports, Montréal. [Baer and al, 989] :Baer, E., Schaffer, J., 989. Soluton mprovement heurstcs for the vehcle routng and schedulng problem wth tme wndows constrants. Amercan Journal of Mathematcs and Management Scences 6 (3,4), [Chang and al, 996] :Chang, W., Russell, R., 996. Smulated annealng metaheurstcs for the vehcle routng problem wth tme wndows.annals ofoperatons Research 63, [Chrstofdes and al, 984] :Chrstofdes, N., Beasley, N., 984. The perod routng problem.networs 4,

208 ANNEX III The Vehcle Routng Problem Wth Tme Wndow [Desrosers and al, 984] :Desrosers, J., Soums, F., Desrosers, M., 984. Routng wth tme wndows by column generaton. Networs 4 (4), [Fsher and al, 994] :Fsher, M., 994. Optmal soluton of vehcle routng problem usng mnmum -trees. Operatons Research 42 (4), [Fsher and al, 997] :Fsher, M., Jornsteen, K., Madsen, O., 997. Vehcle routng wth tme wndows: two optmzaton algorthms. Operatons Research 45 (3), [Holland and al, 975] :Holland, H., 975. Adaptaton n Natural and Artfcal Systems.The Unversty of Mchgan Press, Ann Arbor, MI. [Kohl and al, 997] :Kohl, N., Madsen, O., 997. An optmzaton algorthm for the vehcle routng problem wth tme wndows based on lagrangean relaxaton. Operatons Research 45 (3), [Kolen and al, 987] :Kolen, A., Rnnooy Kan, A., Trenens, H., 987. Vehcle routng problem wth tme wndows. Operaton Research 35, [Osman and al, 993] :Osman, I., 993. Metastrategy smulated annealng and tabu search heurstc algorthms for the vehcle routng problem. Annals of Operatons Research 4, [Potvn and al, 993] :Potvn, J., Rousseau, J., 993. A parallel route buldng algorthm for the vehcle routng and schedulng problem wth tme wndows.european Journal of Operatonal Research 66, [Potvn and al, 995] :Potvn, J., Rousseau, J., 995. An exchange heurstc for routng problems wth tme wndows. Journal of Operatonal Research Socety 46 (2), [Schulze and al, 999] :Schulze, J., Fahle, T., 999. A parallel algorthm for the vehcle routng problem wth tme wndows constrants. Annals of Operatons Research 86, [Thamgah and al, 995] :Thamgah, S., Osman, I., Sun, T., 995.Metaheurstcs for the Vehcle Routng Problem wth Tme Wndows.Techncal Report, SURCpSc- TR-95-32, Artfcal Intellgence and Robotcs Laboratory, Computer Scence Department, Slppery Roc Unversty, PA. 88

209 ANNEX IV Petr Nets ANNEX IV PETRI NETS 89

210 ANNEX IV Petr Nets 90

211 ANNEX IV Petr Nets I. Petr Net: Defntons Petr nets are a graphcal and mathematcal modellng tool used to descrbe and analyse dfferent nds of real systems. Petr nets were frst ntroduced by Carl Adam Petr n 962 n Germany, and evolved as a sutable tool for the study of systems that are concurrent, asynchronous, dstrbuted, parallel and/or stochastc. Performance evaluaton has been a very successful applcaton area of Petr nets. In addton, Petr nets have been successfully used n several areas for the modellng and analyss of dstrbuted-software systems, dstrbuteddatabase systems, flexble manufacturng systems, concurrent and parallel programs and dscrete-event dynamc systems (DEDS) to menton just a few. A mult-agent system s a nd of DEDS that s concurrent, asynchronous, stochastc and dstrbuted. From the DEDS pont of vew, mult-agent systems lac analyss and desgn methodologes. Petr net methods are used n ths wor to develop analytcal methodologes for mult-agent systems. Petr nets are often used n the modellng and analyss of DEDS. They nclude explct condtons under whch an event can occur; capturng also the relatons between concurrent and asynchronous events. As a result, Petr nets are sutable for studyng complex and general DEDS. Ths secton presents an ntroducton to Petr nets. Petr nets are defned followed by mportant propertes and analyss methodologes. Fnally, an example of a manufacturng applcaton s presented. Defnton: The followng s the formal defnton of a Petr. A Petr net s a fve tuple: P, T, A, W, M. where: P s a fnte set of places 0 T s a fnte set of transtons A PT T P s a set of arcs W : A,2,3, s a weght functon P Z M : 0 s the ntal marng 9

212 ANNEX IV Petr Nets The meanngs of places and transtons n Petr nets depend drectly on the modellng approach. When modellng, several nterpretatons can be assgned to places and transtons. For a DEDS a transton s regarded as an event and the places are nterpreted as a condton for an event to occur. Table presents several typcal nterpretatons for transtons and places. Input place Transtons Output places Precondtons Input data Input sgnal Resources needed Condtons Buffers Event Computaton step Sgnal processor Tas or job Clause n logc Processor Post condtons Output data Output sgnal Resource released Concluson Buffer Table : Modellng nterpretatons of transtons and places A smple Petr net example s presented n fgure. Ths example s used later to defne addtonal Petr net characterstcs. Gr t P2 t2 D j G Fgure : Petr net example. Places, transtons and arcs: Places are represented wth crcles and transtons are represented wth bars. The arcs are drected from places to transtons or from transtons to places. The places contan toens that travel through the net dependng on the frng of a transton. A place p s sad to be an nput place to a transton t f an arc s drected from p to t. Smlarly, an output place of t s any place n the net wth an ncomng arc from transton t. In the example (fgure ) p s an nput place of t and p s an output place of 2 t. Transton frng: A transton can fre only f t s enabled. For a transton t to be enabled, all the nput places of t must contan at least one toen (n ths case, t was assumed that the 92

213 ANNEX IV Petr Nets weghts W of the Petr net are equal to one. When the weghts are not ndcated they are assumed to be one. The weght on an arc comng to a transton from one of the ncomng places ndcates the mnmum number of toens needed n the ncomng place n order for that transton to be enabled. When the transton fres, t wll remove from the ncomng place the amount of toens ndcated by the weght of the arc). When a transton s fred, a toen s removed from each nput place, and one toen s added to each output place. In ths way the toens travel through the net dependng on the transtons fred. Defnton 2 (Marng)The marng m of a place p P s a non-negatve quantty representng the number of toens n the place at a gven state of the Petr net. The marng of the Petr net s defned as the functon M : P Z that maps the set of places to the set of non-negatve ntegers. It s also defned as a vector M m m,..., whch represents the j th state of the net. ntal marng s denoted by M. 0 where M, 2 j m p m, M contans the marng of all the places and the j In the example of fgure only transton t s enabled. When t fres, one toen s removed from place p and one toen s added to place p 2. Fgure 2 shows the evoluton of the Petr net n the prevous example. Fgure 2 a) presents the ntal marng of the net M p Mp, Mp 2,0,0 M, 0 2 3, only transton net wth marng,,0 M after t s enabled. Fgure 2 b) presents the t s fred. Here, transtons t and t 2 are enabled and they can be fred. Fnally, fgure 2 c) represents the net after t s fred. In ths case transtons 2 t and 3 t are enabled wth marng,0, M. 2 p t2 P P3 t A r D F t t2 P t P 3 t t P t2 P P t Fgure 2:Petrnet evoluton after frng transtons t and t2. 93

214 ANNEX IV Petr Nets The marng of the Petr net represents the state of the net. As descrbed above, thetranstons change the state of the Petr net n the same way an event changes the state of a DEDS. Defnton 3 (Reachablty graph):the reachablty graph has the marng of the Petr net (or state of the Petr net) as a node. An arc of the graph jonng M wth M represents the j transton when frng taes the Petr net from the marng (state) M to the marng M. j The reachablty graph of the Petr net n fgure s presented n fgure 3. P2 P t2 t3 a) P3 t Fgure 3:Reachablty graph II. Propertes of Petr net: Ths secton covers some of the most mportant propertes of Petr nets such as Reachablty, Lveness, Boundedness and Reversblty. These propertes are essental for the analyss of Petrnet models. Furthermore, they are requred characterstcs for the use of Petr nets nperformance evaluaton. These are propertes that could be appled to mult-agent systems models. Examples of these propertes are boundedness and lveness snce they are related to deadloc avodance n DEDS. Other propertes are gong to be relevant to mult-agent systems partcularly to the communcaton, nteracton, and sngle agent archtectures. 94

215 ANNEX IV Petr Nets. Reachablty: A marng M s sad to be reachable from marng j that taes the Petr Net from state M to M. j M f there exst a sequence of transtons The set of all possble marngs that are reachable from M s called the reachablty set and 0 s defned by R M 0. The reachablty set can be obtaned from the reachablty graph (fgure 3). 2. Lveness: A Petr Net s sad to be lve for a mang 0 M f for any marng n R M 0 t s possble to fre a transton. The lveness property guarantes the absence of dead loc n a Petr Net. Ths property can also be observed from the reachablty graph: f the reachablty graph contans an absorbent state the Petr Net s not lve at that state and t s sad to have a dead loc. If the net s not lve for marng 0 M then at least one marng from R M 0 wll not have any enabled outgong transtons. If the reachablty graph s consdered as the state graph of the net, then an absorbent state s that from whch the marng t s representng does not have any outgong transtons enabled. As a result, when the net reaches an absorbent state, t wll reman n t ndefntely. 3. Boundedness: A Pert Net s sad to be bounded or -bounded f the number of toens n each place does not exceed a fnte number for any marng n R M 0 Furthermore, a Petr Net s structurally bounded f t s bounded for any fnte ntal mang M 0. A Petr Net s sad to be safe f t s -bounded. 95

216 ANNEX IV Petr Nets 4. Reversblty: A Petr Net s reversble f for any mang n RM 0 s reachable. Ths means that the Petr Net can always return to the ntal marng M. 0 For the example n fgure, the reashablty set: R M M,,0, M 0,2,0, M,0,, M 0,,, M 0,0, The Petr net s lve, reversble and 2-bounded for the marng 2,0,0 M. 0. III. Structural analyses Ths secton consders the structural analyss of Petr nets by usng nvarant analyss. Bascally, the lveness and boundedness of the net wll be assessed by usng P-nvarants and T-nvarants. These nvarants are obtaned from the ncdence matrx of the net and they gve nformaton regardng toen conservaton and transton frng sequences that leave the marng of the net unchanged. These concepts are used to assess the overall lveness and boundedness of the net. Defnton (Incdence matrx) let a j w, j t to place p and a j w j, j be the weght of the arc that goes from transton be the weght of the arc from place p to transtont j j. The ncdence matrx A of a Petr net has T number of rows and P number of columns. It s defned as a j A where a a a. j j j The example presented n fgure shows an ordnary Petr net (all the weghts are equal to ) and the followng s ts correspondng ncdence matrx. A Defnton 9 (Net-nvarants) Let A be the ncdence matrx. A P-nvarant s a vector that satsfes the equaton A. x 0 and a T-nvarant s a vector that satsfes the equaton A T. y 0 96

217 ANNEX IV Petr Nets. Boundedness assessment The P-nvarants of the ncdence matrx are used n Theorem to mae an assessment of the boundedness of the Petr net. A Petr net model s covered by P-nvarants f and only f, for each place s n the net, there exsts a postve P-nvarant x such that s 0 x. Theorem A Petr net s structurally bounded f t s covered by P-nvarants and the ntal marng M s fnte Lveness assessment The lveness of the Petr net model s assessed on Theorem 2 by means of the T-nvarants of the ncdence matrx. A Petr net model s covered by T-nvarants f and only f, for each transton t n the net, there exsts a postve T-nvarant y such that y t 0. Ths s a necessary condton but not suffcent. The lveness assessment by the use of T-nvarants s stll an open problem. Theorem 2 A Petr net that s fnte s lve and bounded f t s covered by T-nvarants. 97

218 ANNEX IV Petr Nets 98

219 ANNEX V Introducton to Fuzzy Logc ANNEX V INTRODUCTION TO FUZZY LOGIC 99

220 ANNEX V Introducton to Fuzzy Logc 200

221 ANNEX V Introducton to Fuzzy Logc Fuzzy logc and fuzzy set was ntroduced by Zadeh Lotf as an extenson of classcal set theory, and s bult around the central concept of a fuzzy set membershp functon. Its concept s based on tradng off between sgnfcance and precson. Fuzzy Logc s a convenent way to map an nput space to an output space. Ths concept s used due to ts many advantages, such as, ts naturalness of ts approach and not ts far-reachng complexty, ts flexblty, t s a very powerful tool for dealng qucly and effcently wth mprecson and non-lnearty, t s also tolerant of mprecse data as Fuzzy Reasonng bulds ths understandng nto the process rather than tang t onto the end. As fuzzy logc s nown to deal wth lngustc, vague, and uncertan data, ts use n many applcatons was utlzed to fulfll ths tas. It was cted from the lterature (Martn Hellmann, 200), fuzzy set theory enables the processng of mprecse nformaton by means of membershp functon. In contrast to Boolean Characterstcs Mappng of a classcal set (called crsp set) taes only two values: one, when an element belongs to the set; and zero, when t doesn't. In fuzzy set theory, an element can belong to a fuzzy set wth ts membershp degree rangng from zero set to one. Fuzzy sets are usually dentfed wth these membershp functons as presented n fgure V.. A R Fgure V.: Characterstc Functon of a Crsp Set In addton, basc operatons can be ntroduced on fuzzy sets. Smlar to the operatons on crsp sets, t can be ntersect, unfy and negate fuzzy sets. These operatons concde wth the crsp unfcaton and ntersecton f only the membershp degrees are consdered between 0 and. Examples are shown n (fgures V.2, V.3, V.4 and V.5 f A s a fuzzy nterval between 5 and 8, and B s a fuzzy number about 4. A B x x Fgure V.2: Examples of Fuzzy Set 20

222 ANNEX V Introducton to Fuzzy Logc A B x Fgure V.3: Example of Fuzzy Set between 5 and 8 AND about 4 5 B x Fgure V.4: Example of Fuzzy Set between 5 and 8 OR about 4 A x Fgure V.5: Example of the NEGATION of the Fuzzy Set Fuzzy classfcaton s one applcaton of fuzzy theory. Expert nowledge s used and can be expressed usng lngustc varables (Fgure V.6) 202

223 ANNEX V Introducton to Fuzzy Logc low medum hgh Very low low medum hg 0 90 x 0 x Fgure V.6: Lngustc Varables Fuzzy set theory has also entered a vast doman of applcaton tools, such as fuzzy arthmetc, approxmate reasonng, control, and modelng paradgms. Moreover, n fuzzy rule based systems, nowledge s represented by "IF THEN" rules. Fuzzy rules consst of two parts, an antecedent part statng condtons on the nput varable, and a consequent part descrbng the correspondng values of the output varable. In Mandan type models both antecedent and consequent part consst of fuzzy statements concernng the value of the nvolved varables. Fuzzy rules could be derved from both experts reasonng and lngustc, and from relatonshps between the system varables. There are several defuzzfcaton methods, but the centre-of-gravty formula as llustrated n fgure V.7 s the most frequently used. Also, n order to mprove the model's performance, ts varables and parameters can be adjusted, and the best combnaton can be found by means of smulaton tests. Fnal compute 0 Fgure V.7: Defuzzfcaton usng the Centre of Gravty Approach L.A. Zadeh, Fuzzy sets, Informaton and Control 8 (965)

224 ANNEX V Introducton to Fuzzy Logc J.K. George, Y. Bo, Fuzzy Sets and Fuzzy Logc: Theory and Applcatons, Prentce- Hall, Upper Saddle Rver, NJ, 995. V. Cross, T. Sudamp, Patterns of fuzzy rule-based nference, Internatonal Journal of Approxmate Reasonng () (994) W. Pedrycz, F. Gomde, An Introducton to Fuzzy Sets: Analyss and Desgn, MIT Press, Cambrdge, MA, 998. Bartholo Jr., R. dos S.; Cosenza, C.A.N; DORIA, F.A.M.A.A heurstc algorthm procedure to solve allocaton problems wth fuzzy evaluatons. T Handboo on Reasonng- based Intellgent Systems. 0ed Australa: Lahm c Jan 202, vol.0, p

225 ANNEX VI The Palma de Mallorca Arport ANNEX VI THE PALMA DE MALLORCA AIRPORT 205

226 ANNEX VI The Palma de Mallorca Arport 206

227 ANNEX VI The Palma de Mallorca Arport I. The Palma de Mallorca Arport The Palma de Mallorca Arport (arport code PMI) was orgnally created t handle the sland s postal servce and now over 20 mllon people each year. Known n Englsh as Majorca, Mallorca Arport has one termnal wth four modules, ladled A, B, C and D branchng from t. Although located 8 lometres from the captal. Mallorca Arport s owned by Aena Aeropuertoss. Mallorca Arport has ISO certfcaton as well as contnued nose reducton and nsulaton practces wth surroundng resdental areas. Located n the Medterranean Sea, the sland of Mallorca s the largest of the Balearc Islands and has 550 lometres of coastlne. It receves mllon vstors annually wth Germany accountng for the largest number of travellers whle Span and the UK follow close behnd. Fgure VI. : Palma de Mallorca Arport 207