Seoul Bus Reform and Agenda: Big Data Application for Public Transportation

Transcription

1 Seoul Bus Reform and Agenda: Big Data Application for Public Transportation October 26, 2016 Yongju Yi Keechoo Choi Seong (Jake) Namkoong

2 Contents Ⅰ Ⅱ Seoul Bus Reform in 2004 Some Possible Improvements using Big Data: Issues and Solution focusing on Express Bus Service 2

3 Seoul Bus Reform in

4 Seoul : Geography Russia Area: km 2 Population: 10.3 Million (Metropolitan: 25.5 Million) N. Korea S. Korea Japan China Taiwan 4

5 Hard to Catch Bus (60-70 s) 5

6 Disorder (80-90 s) 6

7 Problems of Bus Operation before Unstable Service by deteriorated bus company 2. Unstable employment 3. Excessive competition to increase revenue 4. Routes owned by private bus company as a patent 7

8 Philosophy 8

9 Measures 9

10 Unified Fare System with Free Transfers up to 5 times Concept of network-wide Integration New fare system Flat Fare Distance-based Fare 10

11 Trunk-Feeder System System Reform by Demand Responsive Operation of 4 types of Bus Interregional lines Feeder lines Trunk lines Circular lines Routes Vehicle Fleet Daily Passengers Million (As of 2013) 11

12 Bus Route Number System Trunk lines Feeder lines Interregional lines Circular lines st digit (1): Departure zone 2 nd digit (4): Destination zone 3 rd digit (2): Route number 1 st digit (2): Departure zone 2 nd digit (4): Destination zone 3 rd, 4 th digit (12): Route number 1 st digit (9): Designated 2 nd digit (4): Departure zone 3 rd, 4 th digit (01): Route number 1 st digit (4): Zone number 2 nd digit (1): Route number 12

13 Exclusive Median Bus Lane Legend Existing After

14 After Initial Exclusive Median Bus Lane System Effect Average Speed: 15km/h 19km/h Punctuality ± 1~2 mins. Daily Passengers 150, Million (2007) 5.75 Million (2013) 14

15 Bus Industry Restructuring semi-public operation system Punctuality 87.3% (2006) 91.4% (2013) Bus Modal Share 25.6% (2003) 27.1% (2013) Satisfaction 59.2 pts (2006) 79.2 pts (2014) 15

16 Bus System Reform Modal Share 45% 40% 35% 30% 25% 20% 15% 10% 5% Bus Subway Taxi Auto Others 0% 16

17 x1000 USD Limitations of Seoul Bus Reform Increase of subsidy to operators costs reduction options needed moral hazard more aggressive tendering system needs 300, , , , ,000 50, Deficits Subsidies Cumulative Deficits (Debt) 17

18 Limitations of Seoul Bus Reform Sustainable fare system VS 18

19 Limitations of Seoul Bus Reform From 6 Elements of BRT Perspectives (USDOT, 2004 * ) Running Ways Fare Collection Partial exclusive median bus lanes Smart card, Pay-on-board Stations ITS Basic shelters BMS/BIS, No priority signal X All same size regular vehicle Vehicles Service & Operating Plans Functional classification of routes * Characteristics of Bus Rapid Transit for Decision-Making, USDOT FTA,

20 BRT in Latin America 20

21 BRT compared between Seoul and Latin America Latin America: Higher BRT standard than Seoul Running ways (exclusive median lane) Seoul Curitiba Mexico City Bogota Quito 119km 70km 67km 105km 37km Stations Outdoor shelter Indoor tube Indoor Platform Vehicles Number of doors in vehicle Regular (10.6m) Articulated (24m) Articulated (18,25m) Articulated (18,24.5m) Articulated (~17m) Fare Collection On-board Pre-board Fare Type Distance-based Flat Number of Corridors Real-time Info. Yes Yes No No 1 of 3 corridors 21

22 Improvements and Solution: Express Bus Service 22

23 Mobility Accessibility and Mobility Trade-off relationship between mobility and accessibility Mobility improvement needs maintaining accessibility? Heavy rail Light rail BRT? Local bus Accessibility Source: TCRP Report 165: TCQSM 3/E (TRB, 2013) 23

24 Mobility Limitation of Bus Transit TT Auto + Parking Time TT Bus + Access/Transfer Time 24

25 Two Alternatives Transfer Madrid The Hague Speed Curitiba Bogota 25

26 Express Transit Services: Typology & Benefits Limited-stop Skip-stop Express Service A (Express) Service B (Local) Service A Service B Service A (Express) (No local service) Others: Zonal, Short-turn, decreased running time decreased fleet size better service (frequency) without new fleet 26

27 Review of Existing Express Services Most routes rapid stops are based on Passenger demand at the stop Opportunity to transfer to other modes/routes Location of passing loop (in case of rail transit) Relative influence on overall commercial speed (e.g. Overpass / underpass availability) Limitations Does not consider O-D patterns Many-to-many or Many-to-one Too much consideration of civil complaints Want to have nearest station as a rapid service stop specific criteria deciding rapid bus stops needed 27

28 Motivations Mostly, existing studies on bus routes planning have limitation in obtaining complete survey data focused on whole network design or large-scale reform Current urban transit network are operating close to optimal solution necessity to improve efficiency without large-scale renovation Urban transportation trend transition in Korea Short-distance in urban area: slow modes (walk, bicycle, tram, ) Mid- or long-distance for inter-city trip: fast modes (BRT, Rapid rail, ) Limitation to improve mobility for inter-city bus route in Korea: narrow stop spacing and highly bended route to maximize revenues increase mobility without decreasing current accessibility 28

29 Attempts to Achieve. Optimizing the express bus service with considering both users and operators Improve competitiveness of public transit on a limited budget while - utilizing smart card data to get base route information and passenger trip data and - developing an optimization model for actual routes 29

30 Smart Card Data More than 98% of passengers usage Tag when boarding and alighting (distance-based fare) Transfers for no additional base fare (up to 5 times, within 30 mins) Data including User class (ordinary, students, senior, disabled, ) Mode codes* (bus, rail, ) Route ID* and Vehicle ID* Total distance(m) Total In-veh time and Total Time (min) Boarding and alighting time* Boarding and alighting stop ID* Total cost * Recording each of 1 st ~5 th used mode 30

31 Smart Card Data Structure User/Mode Route/Veh. ID Boarding/Alighting TIME Boarding/Alighting STOP Ca rd No. No. of Tran sfer Us er Cla ss Mo de Cod e 1 Mo de Cod e 2 Ro ute ID 1 Ro ute ID 2 Veh ID 1 Veh ID 2 Boardin g Time 1 Boardin g Time 2 Alighting Time 1 Alighting Time 2 Boardin g Stop 1 Boardi ng Stop 2 Alighti ng Stop 1 Alighti ng Stop E E E E E E E E E E E E

32 Bus (# M5414) Smart Card Data : Example for Single Travel Ca rd No. No. of Tran sfer Us er Cla ss Mo de Cod e 1 Mo de Cod e 2 Ro ute ID 1 Ro ute ID 2 Veh ID 1 Veh ID 2 Boardin g Time 1 Boardin g Time 2 Alighting Time 1 Alighting Time 2 Boardin g Stop 1 Boardi ng Stop 2 Alighti ng Stop 1 Alighti ng Stop E Subway Line 9 3 #2 Alighting Date/Time: 2013/10/16 08:18:58 Stop ID: 4115 (Yeoui-do Station) Date/Time: 2013/10/16 06:59:44 Mode Code: 570 (M-bus) Route: (M5414) Veh ID: (Veh. Reg. No.) Stop ID: (Gwanggyo-maul) #1 Boarding 1 2 #1 Alighting Date/Time: 2013/10/16 07:56:35 Mode Code: 205 (Subway Line 9) Route: (Subway) Veh ID: (N/A for Subway) Stop ID: 4125 (Shinnonhyun Stn.) Date/Time: 2013/10/16 07:51:41 Stop ID: (Gwanggyo-maul) #2 Boarding 32

33 Smart Card Data Source: ave. statistics per day (Seoul Metropolitan) Taxi, 631 Bus, 5652 Taxi, 5188 Bus, 3818 Subway, 7267 Subway, 5814 Transactions (per 1000) (Total: ) Source: Seoul Government, As of Dec 2014 Paid amount (per 1M KRW) (Total: M KRW 9M GBP) 33

34 Model Development Model inputs Stop-based Hourly OD (Peak, Offpeak) Travel time and distance between stops Vehicle fleet size (local and rapid) Objective function Scenario #1: Minimizing total travel time Scenario #2: Minimizing total costs (user s and operator s) Output Total travel time and total costs Eligible stops for express service Constraint Total number of bus fleet is limited 34

35 Applying Genetic Algorithm Widely-used in TRNDP (Transit Route Network Design Problem) Applying Heuristic methods because of model complexity and huge scale of network Genetic Algorithm: One of the typical global optimum solution finding methods Dimension of solution Solution size: same as the number of stops(stations) (each gene has binary integer whether rapid-bus stop or not at certain station) Expandability for applying real network Huge size, High complexity Expandability for Further research Applying passenger arriving probability, Combined model for route and stops selection model, etc. 35

36 Model function Scenario #1 : Minimizing total travel time N1 N min Z q TT ( i, h, h ) k rs rs n L R r1 sr1 subject to: V V V V L R T L 0, V 0 R i [0,1], i S n n k where hl, hr V V Scenario #2 : Minimizing total costs L R Total travel time of route k where, N : the number of stop of route k, r, s : boarding/alighting stops, q rs : the number of passengers who boarding at stop r and alighting at stop s, TT rs : travel time between r and s (depends on rapid stops set i n and headway h L, h R ) i n : Binary variables for n th stop of Route k stops set S k (Local-only: 0, Local+Rapid: 1) V L, V R, V T : local / rapid/ total vehicles per hour N1 N min Z VOT q TT ( i, h, h ) VOC V VST k rs rs n L R R k r1 sr1 subject to: V V V V L R T L 0, V 0 R i [0,1], i S n n k where hl, hr V V L R where, VOT : Value of time (passenger) VOC : Vehicle operating cost per hour VST k : Rapid vehicle saved time N1 t (1 i ) n1 reduced n, n1 n 36

37 Four types of OD influencing TTrs Rapid bus stops at boarding stop (r) Rapid bus stops at alighting stop (s) Available bus service Type 1 Local / Rapid Type 2 Local / Rapid + Local Type 3 Local / Local + Rapid Type 4 Local / Local + Rapid + Local Rapid Service (R) Local Service (L) Stop No. Example path of Type 1 (R R) Example path of Type 2 (R L) Example path of Type 3 (L R) Example path of Type 4 (L L) 37

38 In-vehicle travel time of rapid and local bus s1 s1 R reduced trs tn, n1 tn, n1 (1 i n) nr nr Rapid bus travel time between r and s (reduced when rapid bus passes stop n+1) t L rs s1 nr t n, n1 Local bus travel time between r and s Reduced time by skipping w t AccDcc OnOffTime C (1 g/ C) C ((1 g/ C) C) AccDcc OnOffTime n reduced w1 w1 n, n1 n Acceleration + Deceleration time saving (AccDcc): 11.6 sec per stop (Robinson, 2013) Boarding and alighting time saving (OnOffTime j ) Boarding: 2.3sec/pax, Alighting: 2.0sec/pax (smartcard data) OnOffTime n Intersection signal delay saving Average cycle C: 162 sec, Average g/c: 0.42 (In case of Suwon city) Arriving intersection faster as amount of AccDcc and OnOffTime n Reducing expected delay time (go through green light without stopping) C N N q q VL VR in VL VR in ns rn s1 r1 max 2.3, 2.0 w Intersection signal delay saving (details on the below) 38

39 Total Travel Time (r to s) by four service types Local bus Rapid bus Local + Rapid bus Rapid + Local bus s1 L hl L TT w t OnOffTime 2 rs wait rs n nr s1 R hr R TTrs wwait trs OnOffTime n (1 in ) 2 nr a1 s1 LR hl hr L R TTrs wwait wtransfer tra tas OnOffTime n OnOffTime n (1 in ) 2 2 RL hr hl TTrs wwait wtransfer t 2 2 nr where, w wait : Waiting time weight w transfer : Transfer time weight h L : headway of local bus h R : headway of rapid bus na a1 s1 R L ra tas OnOffTime n (1 in ) OnOffTime n nr na Local + Rapid + Local bus a1 LRL hr hr hl L R L TT w w t t t OnOffTime nr rs wait transfer ra ab bs n b1 s1 OnOffTime (1 i ) OnOffTime na n n n nb 39

40 Changing origin or destination stops Changing stops when total travel time is reduced Walk accessibility of bus stop: 500m (average of preceding research) For the passengers that using local stop within 1km of express bus stop, he/she change there stop to express if total travel time is reduced Rapid bus stops at boarding stop (r) Rapid bus stops at alighting stop (s) Available bus service Type 2 Local / Rapid + Local / Rapid + Walk Type 3 Local / Local + Rapid / Walk + Rapid Type 4 Ratio of transferred OD Changing ratio from local-only stop: x/(x+y) in below figure Avg. Walking Distance x y Local / Local + Rapid + Local / Local + Rapid + Walk / Walk + Rapid + Local Express Stop 500m Local-only Stop Before After Boundary of each stops 40

41 Fitness Evaluation Process (in case of minimizing total costs) OVTT weight Stop to Stop Travel Time/Dist. Local/Rapid headway Peak hourly O/D Offpeak hourly O/D Rapid Stops Set Stop change (Local Rapid) Stop-to-stop travel time Headway Travel time comparison whether origin / dest. is changed or not (by OD types) modified OD, additional access time i+1th generation Solution Pop. (Genetic Algorithm) Total travel time estimation model by OD types Reduced run time estimation model by OD types Total travel time Total reduced run time Passenger VOT (Value of Time) Transport Cost Total social cost No Convergence? Yes Final solution (Rapid stops Set) <Legend> Given Data Est. Model Input Data Model Output 41

42 Case study Anyang Uiwang Route # 300 Trunk line Round 72.9km Ajou Univ. Suwon 12,438pax/day (2013) Peak 1,141pvh Offpeak 525pvh 139 Stops (round) Osan headway 6~8 min Avg. Trip distance 9.8km 42

43 Results (Scenario #1: Minimizing total travel time) iterations (MATLAB) Total travel time is decreased as much as 4,362 min Rapid* (vph) Local* (vph) # of Rapid stops Total travel time (min) Travel time per passenger (min) Total Social Cost per day (KRW)** Rapid Bus Usage Rate (%) Travel time per passenger (min) when increasing allocation*** of , ,196, of , ,327, of , ,779, of , ,209, of , ,756, of , ,634, of , ,488, of , ,559, of , ,004, of , ,073, of , ,957, * As of peak hour (offpeak hour: 75% of peak hour) ** VOT: 3,620KRW/hour, VOC: 35,415KRW/veh/hour (as of 2013, source: Korea Development Institute and City of Seoul govt.) *** Increasing local buses allocation under total social cost does not exceed as amount of the base case (Local bus service only) 43

44 Results (Scenario #2: Minimizing total costs) iterations (MATLAB) Total cost is decreased as much as 3.2 million KRW (2,750 USD) per day with 1.1 min increasing of travel time per passenger Rapid* (vph) Local* (vph) # of Rapid stops Total travel time (min) Travel time per passenger (min) Total Social Cost per day (KRW)** Rapid Bus Usage Rate (%) of , ,196, of , ,175, of , ,570, of , ,932, of , ,416, of , ,264, of , ,955, of , ,747, of , ,786, of , ,965, of , ,651,

45 Results of rapid service stop selection Scenario #1 Scenario #2 Rapid stop Rapid skip Rapid stop Rapid skip 45

46 Model Verification Comparing with existing stop selection criteria Selecting the top quartile of stops with the most activity (O/D) for users of target route (Tetreault and El-Geneidy, 2010) 35 stops (25%) are selected from total 139 stops Rapid stop Rapid skip 46

47 Results The number of rapid stops is fixed regardless of rapid/local fleet size Travel time is increased when introducing express service Total costs are reduced, however smaller than our developed model Rapid* (vph) Local* (vph) # of Rapid stops Total travel time (min) Travel time per passenger (min) Total Social Cost per day (KRW)** Rapid Bus Usage Rate (%) of , ,196, of , ,684, of , ,301, of , ,909, of , ,547, of , ,491, of , ,195, of , ,041, of , ,225, of , ,371, of , ,484, * As of peak hour (offpeak hour: 75% of peak hour) ** VOT: 3,620KRW/hour, VOC: 35,415KRW/veh/hour (as of 2013, source: Korea Development Institute and City of Seoul govt.) 47

48 Results Comparisons Rapid* (vph) Local* (vph) Tetreault and El-Geneidy, 2010 Model (Scn. #1) Model (Scn. #2) # of Rapid stops Travel time per passenger (min) Total Social Cost per day (KRW)** # of Rapid stops Travel time per passenger (min) # of Rapid stops Total Social Cost per day (KRW)** of ,196,652 0 of of ,196, of ,684, of of ,175, of ,301, of of ,570, of ,909, of of ,932, of ,547, of of ,416, of ,491, of of ,264, of ,195, of of ,955, of ,041, of of ,747, of ,225, of of ,786, of ,371, of of ,965, of ,484, of of ,651,200 * As of peak hour (offpeak hour: 75% of peak hour) ** VOT: 3,620KRW/hour, VOC: 35,415KRW/veh/hour (as of 2013, source: Korea Development Institute and City of Seoul govt.) 48

49 Conclusions Seoul Bus Reform 2004 Half success need to address more issues like transfer, overtaking, station amenity, full fledged BRT making, and pricing Improving bus services Transfer (Infrastructure) Rapid Services (Operational approach) Limited-stop bus stop optimization model Considering several constraints and options Improve competitiveness of transit in a limited budget Improve operating efficiency of transit routes 49

50 Thank you 50