[EN-037] Airborne Conflict Modeling and Resolution for UAS Insertion in Civil Non-Segregated Airspace

Transcription

1 ENRI Int. Worksho on ATM/CNS. Tokyo, Jaan. (EIWAC 2010) [EN-037] Airborne Conflict Modeling and Resolution for UAS Insertion in Civil Non-Segregated Airsace (EIWAC 2010) C.A. Persiani*, S. Bagassi** *Deartment of Electronic, Comuter Sciences and Systems- University of Bologna **Deartment of Mechanical, Nuclear, Aviation and Metallurgical Engineering - University of Bologna Abstract: In this aer we resent a Genetic Algorithm designed to manage the mission of an UAS that has to visit a set of mission oints into a congested airort TMA (Terminal Area). The genetic aroach is useful to model the resence of different avoidance otions: oulations of ilots that have different avoidance hilosohy are crossed in order to obtain a good mix of avoidance technique. Our methodology is based on a strategic, intent-based, non-cooerative (only one aircraft the UAV maneuver), geometric (rediction is based on geometrical rojections) and distributed (as oosed to centralized) aroach. Finally, real iloted traffic data from the Milano Linate (LIML) Terminal Area are used to test the algorithm. Keywords: Airborne conflict resolution, UAS insertion 1. I TRODUCTIO One of the most interesting challenge of the next years will be the Air Sace Systems automation. This rocess will involve different asects as the Air Traffic Management, the Aircrafts and Airort Oerations and the Guidance and Navigation Systems. The use of UAS (Uninhabited Aerial System) for civil mission will be one of the most imortant stes in this automation rocess. The UAS insertion in civil non-segregated airsace addresses quite a number of issues such as the level of riority that unmanned air traffic should have relative to manned air traffic, the coordination between the UAV controller and the air traffic control, the level of automation of the UAS, the sense & avoid caacity of the UAS. The management of otential conflicts between UAS and manned air traffic is a challenging issue that is often considered as an instance of the general Conflict Detection & Resolution (CD&R) roblem. Since the first concets concerning Free Flight were envisaged, both the industry and the academia research communities aid attention to CD&R roblem. In order to study the feasibility of self searation, many rototye tools such as the Autonomous Oerations Planner (AOP), Future ATM Concets Evaluation Tool (FACET) [1] develoed at NASA and the Airborne Searation Assurance System (ASAS) [2] develoed at the National Aerosace Laboratory (NLR) in the Netherlands have been roosed. All these tools imlement CD&R algorithms. CD&R algorithms can be airborne or centralized. Usually airborne conflict detection and resolution algorithms are suitable for tactical use, whereas centralized CD&R algorithms can be used both for tactical and for strategic control. CD&R algorithms can be state-based or intentbased. State-based refers to the use of aircraft state information as oosed to those algorithms that use intent information, e.g. flight lan. The airborne algorithms used in the Autonomous Mediterranen Free Flight (AMFF) simulations have roved to be effective with limitations for dense traffic conditions [3]. AMFF is a state-based conflict detection and resolution concet with a level of automation such that the ilot follows the automatically generated conflict resolution advices by steering the aircraft. Farley et al. [4] assessed the erformance of a conflict resolution algorithm develoed as art of the Automated 123

2 C.A. Persiani, S.Bagassi Airsace Concet [5] for conflicts detected in any hase of flight - including arrival merging oerations - in a simulation environment designed to reresent the comlexity, variety and volume of current and future air traffic oerations. Afterwards, the literature has exanded to include aroaches based on genetic algorithms [6][7][8] and arrival-time constraints. In this aer we resent a Genetic Algorithm designed to manage the mission of an UAS that has to visit a set of mission oints into a congested airort TMA (Terminal Area). The genetic aroach is useful to model the resence of different avoidance otions. Our methodology is based on a strategic, intent-based, non-cooerative (only one aircraft the UAV maneuver), geometric (rediction is based on geometrical rojections) and distributed (as oosed to centralized) aroach. It is comrised by two stes. In a first ste a geometric analysis allows to identify all ossible UAS aths among the mission targets and their related otential conflicts with the iloted aircraft (traffic aircraft). Air Traffic Management oerative techniques are used to model different otions of conflicts resolution: vertical and horizontal avoidance, seed regulation and the use of the holding atterns. These avoidance otions are then comared taking into account the mission constraints and objectives, minimum time or minimum fuel, in order to define a cost for each UAS ath and its related conflict avoidance otions. In the second ste the roblem of find the mission targets visit order is considered. It consists in a combinatorial roblem that concerns the sequencing of both targets and conflict resolution otions. Finally, real iloted traffic data from the Milano Linate (LIML) Terminal Area are used to test the algorithm. 2. PROBLEM FORMULATIO Let us consider a UAV mission erformed in Controlled Air Sace. The mission consists in dearting from a ground base, taking ictures over a set M w (Mission Way Points) of m targets at a given mission altitude and coming back to the same airfield. During this mission a set A of n iloted aircrafts fly into the same Air Sace following authorized routes: the UAV has to maintain a minimum searation from the iloted air traffic. We suose that the osition and the altitude of each aircraft during the entire mission (off-line aroach) are known. Currently ICAO regulations concerning minimum searation between iloted and not iloted traffic do not exist: we assume a minimum longitudinal searation of 5 NM and a minimum vertical searation of 1000 ft. According to its erformance, the UAV has different otions to avoid the iloted air traffic: holding over a way oint, reducing or increasing its seed or using a vertical or horizontal avoidance. The goal consists in lanning the UAV minimum-time and minimum-fuel routes ( in terms of succession of mission targets) that allow to visit the targets maintaining the minimum searation from the iloted traffic. This roblem has two decisional dimensions: the choice of the best order of visiting a set of way oints and the choice of the avoidance manoeuvres that have to be used. The first roblem can be considered as the well known Travelling Salesman Problem (TSP), while the second is ertaining to an UAV ath lanning roblem that has been investigated using several aroaches. These two roblems can be joint considering that the avoidance of the iloted air traffic in a sub-ath between two mission way oints deends on the moment by which the UAV starts flying between this coule of way oints. In this way it is ossible to formulate the roblem as a Time-Deendant Travelling Salesman Problem (TDTSP). The temoral dimension can be introduced by dividing into time stes the temoral horizon. If t end is an uer bound on the mission duration obtained by the UAV endurance and t is the duration of a time ste, the set T = { 1,... h} is the set of the time stes where h= tend / t reresents the number of time stes used. For a mathematical formulation of this roblem see [9]. Considering long endurance missions (7/8 hours and u to 40/50 targets) and different avoidance otions, the roblem calls for heuristic solution. Our aroach uses a first rerocessing ste by which the function of the time w( k i, j) ( t) is calculated for each UAV sub-ath: it reresents the weight (as estimation of the time and fuel consumtion ) of the sub-ath (i,j)started by the UAV at time ste t T using the k avoidance manoeuvre. These weights are then used by the genetic algorithm to calculate the oulation fitness. Poulations of ilots that have different avoidance hilosohy are crossed in order to obtain a good mix of avoidance technique. 3. CO FLICT GRAPH A D AVOIDA CE MA OEUVRES As the TSP is associated to a grah, the UAV mission environment can be modeled using a Conflict Grah. We define a Conflict Grah a sixtule G= { V, M w, A, C, K, T}. 124

3 ENRI Int. Worksho on ATM/CNS. Tokyo, Jaan. (EIWAC 2010) V = {( i, j) / i M w, j M w, i j} is the set of ossible aths between the mission way oints: if no conflict occurs, the UAV flies direct from i-th to j-th target. Due to the different altitude values required to overfly the targets, the ath (i,j) is different from the ath (j,i) in terms of time and fuel consumtion: the Conflict Grah is an asymmetric and direct grah. C is the set of ossible conflicts: a conflict c C between the UAV u and an aircraft a A occurs if the minimum searation requirement is not satisfied during all the time stes t T. K is the set of all ossible avoidance otions: holding, seed control and avoidance on vertical and horizontal lanes. In order to build the Conflict Grah it is necessary to consider the intersection between the UAV Mission Grah G m = { V, M w} and the route of each aircraft a A. Such routes consist in a set of IFR (Instrumental Flight Rules) oints (RadioAssistance or Fix Points): a A, R a = { a,... }, n a. As inut of the roblem (off-line 1 a n a roblem) the altitude h a a i and the time a t of the aircraft a a i over its routing oint a i are known for each aircraft a A. An intersection between a UAV ath (i,j) and an aircraft sub-route (a i,a j ) occurs if the distance between the closest oints of these segments is less than 1000 ft on the vertical lane or 5 NM on the horizontal lane. In this case a conflict c C in the ath (i,j) and its related time ste t c are identified. The weights (in terms of time and fuel consumtion) of the avoidance manoeuvres of the conflict c deends on: the dearting time ste from i to j; the geometry of the conflict; the UAV erformances. Naturally, starting a ath in a secific time ste and maintaining a defined seed could imlicate a conflict while another dearting time ste could be conflict free. The fuel and the time consumtion function of the dearting time ste are estimated using, for each UAV ath, the geometric model described below Holding An holding is an avoidance manoeuvre that consists in erforming an orbit over the initial way oint i of the ath (i,j). The manoeuvre is comrised by an in-bound leg and by an out-bound leg. An holding over a target oint i can be modeled by forbidding the dearting from that oint for i i i a subset of time stes T f = [ thstart... thend ] T i f T, where i t hstart is the time ste of holding start and i t hend is the time ste of holding end. For given UAV erformances, these time stes are determined considering the geometry of the conflict c and the related time ste t c. Let us consider Fig.1 where an intersection between the UAV ath 1u - 2u and the aircraft a sub-route 1a - 2a occurs in the oint cnf. Figure 1. A conflict between a UAV ath and an aircraft sub-route defines a conflict zone and its related characteristic oints. The conflict time ste t c is the overflying time ste of the oint cnf. A sub-searation stretch over 1u - 2u is identified between 1 c u and 2 c u. We define the oint 1 c u as the first oint over the UAV ath between 1u and cnf that resents a distance of 1000 ft on the vertical direction or 5NM on the horizontal direction, whichever haens first, moving from cnf to 1u. Analogously the oint 2 c u is the first oint over the UAV ath between cnf and 1u that resents a distance of 1000 ft on the vertical direction or 5NM on the horizontal direction, whichever haens first, moving from cnf to 2u. In the same way the oints 1 c a and 2 c a are defined over the aircraft a route. The oints 1 c u, 2 c u, 1 c a, 2 c a define a sub searation zone around the conflict oint cnf. Considering a fixed seed for the UAV v uav (for examle the best range seed, the best climb seed or the best glide seed), its last dearting time ste from the target oint 1 u is such that it can arrive in 2 c u when the aircraft a is in the oint 1 c a. If is the time ste interval necessary for the UAV t 1u 1c u to fly from 1u to 1 c u maintaining v uav, the holding over 1u starts at time ste: 1u thstart = max{ 0, t a + t _ c t _ 2c 1 1a 1 a 1u Analogously the first UAV restarting time ste from the target oint 1 u is such that it can arrive in 1 c u when the aircraft a is in the oint 2 c a. The holding over 1u finishes at time ste: 1u thend = max{0, t a + t c t c 1 1a _ 2 a 1u _ 1 u u } } 125

4 C.A. Persiani, S.Bagassi The holding over 2u is determined in the same way considering the UAV flying from 2u to 1u. 2u thstart = max{0, t a + t 2u thend = max{0, t a + t c t c 1 1a _ 1 a 2u _ 1 u c t c 1 1a _ 2 a 2u _ 2 u } } examle, let us consider the case reresented in Fig.4, when the UAV flies from oints 1u to oints 2u. The direction of the UAV after 2u is not known, so it is necessary to consider a safe buffer zone through the construction of a shere around cnf. Whatever the UAV direction is, in the subsequent ath the holding over 1u allows to avoid the mentioned conflict. Moreover, once a conflict free time ste for the dearture from the ground base has been identified and this rocedure is reeated for each conflict c C for the all UAV ath ( i, j) V, the dearture from 2u to 1u cannot haen during a conflict. The intersection case resented in Fig.1 is the general case by which several articular cases can be derived. Fig.2 reresents an intersection between the UAV ath 1u - 2u and an aircraft route such that more than one aircraft sub-route is involved: the aircraft direction change in the sub searation zone. Figure 4 The conflict zone involves more than one UAV ath Finally for each dearting time ste it is now known the t related holding h i, j to reach the next target oint avoiding the conflicting traffic. If the UAV is flying maintaining the seed v, it is ossible to estimate the weight of the UAV sub-ath (i,j) as: w holding ( i, j) ( t) t ( h, ( 1u _ 2u / v)) = i j + Figure 2 and Figure 3. The conflict zone involves more the one aircraft sub route The oint 2 c u is defined through the distance from the sub-route 2u - 3u and the UAV ath 1u - 2u ; moreover the oint 2 c a is on the sub-route 2u - 3u. In case the oint 2 c u 1 u is between 1u and 1 c u,then thstart is defined using the * aroximation of the oint 2 c u that corresonds to the rojection of 2u over the UAV ath. The same haens for the oint 1 c u if the direction change occurs as in Fig.3. In case the oint 1u or 2u is in the sub-searation area as in Fig.4, the oints 1 c u and 2 c u are defined using a safe shere with radius of 5NM and center in cnf. For 3.2. Seed Control The seed control is often used by the Air Traffic Control to maintain or to rovide a minimum searation between aircraft of different erformances without changing their routes. Using the same geometric model resented for the holding, red it is ossible to find the medium seeds v m and v inc m that the UAV has to maintain to avoid a conflict by reducing or increasing resectively its medium seed. Considering only the case reorted in Fig.1, v red m is the medium seed that the UAV has to maintain to arrive in 1 c u when the aircraft is in 2 c a starting the ath 1u - 2u at time ste t 1u : red v m ( t 1u ) =( 1u_ 1 c u )/( t 1a + t 1a _ 2ca ) 126

5 ENRI Int. Worksho on ATM/CNS. Tokyo, Jaan. (EIWAC 2010) Similarly, v inc m ( t 1u ) is the medium seed that the UAV have to maintain to arrive in 2 c u when the aircraft is in 1 c a : inc v m ( t 1u ) =( 1u_ c )/( 2 u 1a + t a ca ) t 1 _ 1 Once identified the medium seeds, it is necessary to analyze if they are consistent with the UAV erformances: both seeds must be smallest or equal to the UAV maximum seed and greater than the UAV stall seed. Furthermore the thrust available should allow the UAV to reach the calculated seed in a manner consistent with the conflict: we suose the acceleration/deceleration hase is consistent with the conflict. If it is ossible to reduce or increase the UAV seed, the estimation of the weight of the ath (i,j) due to seed control w scontr ( i, j) ( t) can be written as: scontr inc / red w( i, j) ( t) = ( 1u _ 2u )/ vm ( t) 3.3. Horizontal and vertical avoidance Several models have been develoed for the horizontal or for the vertical avoidance. MILP techniques, for examle, allow to obtain an otimal trajectory that avoids a fixed or moving obstacle by modeling an exclusion zone around it. Instead our aroach for the horizontal avoidance is based on Air Traffic Management oerative technique that uses vectoring to avoid collision between converging traffic. This manoeuvre is reresented in Fig.5: the aircraft A is vectored and directed toward the aircraft B. Oening its route, the aircraft A reaches the osition of B when B will be in a safe osition. Naturally this technique is alicable only for articular convergence angles between the aths. In Fig.6 the manoeuvre is modeled using our geometric aroach. A safe shere is built around the osition of the aircraft a at each time ste starting from t 1u. The vectoring is modeled considering the tangent to the safe shere that asses through the current osition of the UAV and behind the aircraft osition. Figure 6. The geometric aroach used to model ATC avoidance technique The osition of the UAV is udated for the next time ste by a shift of v t. Once the UAV assed behind the aircraft it may roceed direct to the oint 2u. If n v is the number of time stes required to ass behind the aircraft and n is the number of time stes required to roceed back to the original route to oint 2u, it is ossible to estimate the weight of the UAV subath (i,j) as w( Havoid i, j) ( t) = nv t+ n t Vavoid w i, j ) The vertical avoidance is modeled creating an alternative UAV sub-route over or under the aircraft route. The model detects ossible further conflicts of the alternative subroute. The weight ( of the ath (i,j) is estimated directly from the UAV erformances (rate of climb, max efficient seed, rate of descent) and from the length of the alternative generated ath. 4. CO FLICT DETECTIO ALGORITHM Figure 5 : ATC avoidance technique on the horizontal lane: the aircraft A is vectored to the osition of B in order to ass behind it In the following seudo-code, we resent a simle geometric algorithm (with comlexity O ( A V ) ) based on the Conflict Grah that allows to comute the weight w k t ( i, j ) ( ) of the UAV ath (i,j) as a function of the 127

6 C.A. Persiani, S.Bagassi dearting time ste from i. The conflict condition between the UAV ath (i,j) and the aircraft a route, considering a UAV seed v, is indicated as c( Ra, ( i, j), v) selected in order to obtain a good ilot for a secific route. Finally, the genetic algorithm resents fast comutation time comatible with future real time alications: if the comutation time is smaller than the time ste, a relanning can be done at each time ste considering tactical udating. Using this model, the resence of a conflict on a ath (i,j) is related to the dearting time ste: if a conflict occurs at seed v, the arches weight deends on the avoidance otion and it is calculated by the model resented. Instead, if no conflict occurs, the arches weight corresonds to the flight time, considering the Euclidean distance at seed v. The UAV ath weight allows to rovide an estimation of the mission endurance, while to calculate the fuel consumtion it is necessary to use the secific look-u table of the UAV. These table rovides the fuel consumtion er time unit as a function of the TAS and of the altitude. 5. SOLUTIO APPROACH: GE ETIC ALGORITHM The Travelling Salesman Problem has been hardly discussed in ast years: many exact and heuristic aroaches have been roosed in literature. Instead researches on the TDTSP (Time Deendant TSP) are quite rare; interesting algorithms have been roosed to deal with traffic jam. Frequently, algorithms that rovide good solution for the TSP do not allow to obtain valid results in the TDTSP. For examle, let us consider a solution that resents crossing arches: it is non-efficient circuit for the TSP. Instead that solution could be valid for the TDTSP. Moreover the roblem we are dealing with requires the choice of the avoidance techniques couled with the choice of the ath. In this context, a genetic algorithm reresents a valid aroach. Firstly, it allows a better management of the time deendence than other heuristic aroaches. In fact, while the fitness of a route is time deendent, the oerations of chromosome recombination are not time deendent. This allows to by-ass several roblems tyical of some constructive heuristics in the TDTSP. Moreover, the evolutionary rocess can easily be extended to the choice of the avoidance technique: different Avoidance Philosohy can be roerly combined and The algorithm we roose starts randomly generating a first oulation of η UAV routes R u. At each route a Pilot Philosohy k r is randomly assigned. It consists in one of the avoidance techniques k K resented: holding(k=0), seed control (seed reduction k=1, seed increasing k=2), avoidance on vertical lane (k=3), avoidance on horizontal lane (k=4). In each route of the first oulation, the UAV uses only one avoidance technique to solve all the conflicts: it looks like it is managed by a not efficient ilot that is able to avoid the traffic only in one way. The fitness (min time or min fuel) of each route is calculated considering the weight of the arches function of the dearting time ste (calculated trough the geometric re-rocessing algorithm resented) and the oulation is sorted by decreasing fitness. Then the evolution to the next oulation starts. This rocess is reeated until the STOP criterion is satisfied. It consists in reeating η t times this evolution, checking if the imrovement in the last η c individual is greater than 1%. If that imrovement is less than 1% the evolution is terminated. During the evolution different Pilot Philosohies are combined each other as well as the aths of the routes are combined. The evolution rocess selects the most efficient ilot for a secific route. The evolution to the following oulations is based on cross over oerations alied to the array used to store the route and the avoidances manoeuvres. The first half of the oulation, sorted by decreasing fitness, is considered good : the second art of the oulation is built from the chromosome of the revious art. The oerations used are random shuffle, rotate and reverse as available in the Evolutionary Concert Tour (ECT) for C++ Builder 6 ( Note that this evolutions oeration are not alied to the first and to the last element of the array because they corresond to the dearting/arriving aerodrome. The Pilot Philosohies associated to the ath are crossed too and the ath obtained resents a better mix of avoidance otions. Finally the oulation obtained is sorted again by 128

7 ENRI Int. Worksho on ATM/CNS. Tokyo, Jaan. (EIWAC 2010) decreasing fitness. The genetic algorithm used is described in the following seudocode where the generic term route indicates the combine of a set of aths and a set of avoidances: green one; it is also ossible to recognize the navigation oints and the aircraft indicated with red squares. BEGIN 1) Prerocessing: comutation of w(t.); 2) Generation of the first oulation of η UAV routes and assignment of the Pilot Philosohy k r ; 3) Sort the oulation by decreasing fitness (min time/min fuel);generation=1; STOP=FALSE 4) While STOP==FALSE begin Evolve oulation: first η /2 routes in solution; second η /2 routes crossed; Sort the oulation by decreasing fitness; generation=generation+1; if(generation<=η t )or(imrovement(η c )<=0.01) then STOP=TRUE; end END 6. SIMULATIO RESULTS We test our aroach on a real air traffic scenario: the TMA (Terminal Manoeuvring Area) of Milano Linate (ICAO code LIML), a major airort in the North of Italy with an average of 350 movement (air) er day. In this area Navigation Points as Radio-Assistance and Fix Points (radial and distance by a radio assistance) are reorted; SID (Standard Instrumental Dearture) route and STAR (STandard arrival Route) of the airort are modeled using grahic tools. Air traffic data related to the day of November 12 th 2009 are acquired by AOIS (Aeronautical Oerational Information Sistem) and Radar Track rovided by ENAV S..A (Italian Agency for Air Navigation Services). Position and altitude of 115 aircraft (arrival, dearture and overflying traffic) are simulated starting from 5:30 UTC (Universal Time Coordinated) for 6 hours of simulation. We generate the UAV target oints considering that some of the UAV target oints corresond to the aircraft navigation oints, the others have been randomly generated. In Fig.7 an examle of mission simulation is reorted. The UAV route is the red one while the aircraft routes are the Figure 7 Examle of results of the simulation with 20 targets. The UAV route is red, the aircraft routes are green. An interesting examle of UAV conflict resolution is reorted in Fig 8, 9, 10,11, that shall to be observed subsequently. The oints AMOXI, LIMBA and DIXER are lined u to Runway 36, arriving aircraft coming from South follow this route. Looking the ictures in sequence it is ossible to recognize an arrival sequence AZA2036, ACL324 and AZA2032. The UAV route includes the ath between DIXER and LIMBA in oosite direction. However, this ath is erformed by the UAV between ACL324 and AZA2032 without searation minima infringement. 129

8 C.A. Persiani, S.Bagassi Figure 8, Figure 9, Figure 10, Figure 11. Examles of UAV route management in a critical area, the long final or ILS ath: the UAV uses this ath between sequenced aircraft without infringing the minimum searation. Finally, in the tables below simulation results are reorted. They consider different mission dimensions (10, 20, 25, 30, 35, 40 and 45 target oints) for both objectives of minimum fuel (Table 1) and minimum time (Table 2). The results rovided consider the following arameters: η = 200, η t = 3500, η c = 330, and t = 1 minute, comuted on an IntelCore Duo 2GHz

9 ENRI Int. Worksho on ATM/CNS. Tokyo, Jaan. (EIWAC 2010) Obj Value N of Conflicts Table 1 Simulation results with minimum fuel objective Holding Seed Red. Seed Inc. H Avoidance V Avoidance CPU time 20 targets ,78 25 targets ,12 30 targets ,41 35 targets ,66 40 targets ,47 45 targets ,98 Obj Value N of Conflicts Table 2 Simulation results with minimum time objective Holding Seed Red. Seed Inc. H Avoidance V Avoidance CPU time 20 targets ,83 25 targets ,67 30 targets ,69 35 targets ,8 40 targets ,48 45 targets ,53 7. CO CLUSIO S We resent the roblem of the management of an UAV mission into controlled air sace. In this aer the roblem has been formalized as a TDTSP then a geometric model has been rovided to calculate the weight of the UAV aths as a function of time. Finally a genetic algorithm has been resented to solve the roblem in real traffic scenarios. Simulation results show how the roosed geometric model efficiently defines the arches weights to be used in the conflict resolution. The genetic algorithm allows to deal efficiently with the roblem s time deendence, moreover it is useful to identify a roer sequence of avoidance maoeuvres. Finally comutation time shows that this aroach could be alied to future real time alications. 8. REFERE CES [1] K. Bilimoria, B. Sridhar, G. Chatterji, K. Sheth, and S. Grabbe, FACET: Future ATM concets evaluation tool, in 3rd USA/Euroe Air Traffic Management R&D Seminar, Nales, Italy, June [2] J. Hoekstra, R. Ruigrok, R. van Gent, J. Visser, B. Gijsbers, M. Valenti, W. Heesbeen, B. Hilburn, J. Groeneweg, and F. Bussink, Overview of NLR free flight roject , Tech. Reort NLR- CR , National Aerosace Laboratory (NLR), May [3] Blom, H.A.P., Obbink, B.K. and Bakker, G.J., Safety Risk Simulation of an Airborne Self Searation Concet of Oeration. 7th AIAA- ATIO Conference, Setember 18-20, 2007, Belfast, Northern Ireland. [4] Farley, T., Kufer, M., Erzberger, H., Automated Conflict Resolution: A Simulation Evaluation Under High Demand Including Merging Arrivals, 7th AIAA-ATIO Conference, Setember 18-20, 2007, Belfast, Northern Ireland. [5] Erzberger, H., The Automated Airsace Concet, Proceedings of the 4th USA/Euroe Air Traffic Management R&D Seminar, Santa Fe, New Mexico,

10 C.A. Persiani, S.Bagassi [6] Mondoloni, S., Conway, S., Airborne Conflict Resolution for Flow-Restricted Transition Airsace, AIAA , American Institute of Aeronautics & Astronautics, Reston, Virginia, [7] Vivona, R. A., Karr, D. A., Roscoe, D. A., Pattern-Based Genetic Algorithm for Airborne Conflict Resolution, AIAA , American Institute of Aeronautics & Astronautics, Reston, Virginia, [8] Karr, D.A., Vivona, R. A., Roscoe, D.A., Consiglio, M., Exerimental Performance of a Genetic Algorithm for Airborne Strategic Conflict Resolution, 8th USA/Euroe Air Traffic Management Research and Develoment Seminar (ATM2009). [9] Bagassi S., Bombardi T., Francia D., Persiani C.A., 3D Trajectory Otimization for UAS Insertion in Civil Non-Segregated Airsace, AIAA , AIAA Modeling and Simulation Technologies Conference, Chicago, Illinois, Aug , Coyright Statement The authors confirm that they, and/or their comany or institution, hold coyright of all original material included in their aer. They also confirm they have obtained ermission, from the coyright holder of any third arty material included in their aer, to ublish it as art of their aer. The authors grant full ermission for the ublication and distribution of their aer as art of the EIWAC2010 roceedings or as individual off-rints from the roceedings