Webinar: Modelling mode and route choices on public transport systems

965 views
744 views

Published on

2013-12-05 Webinar by Sebastian Raveau

Published in: Technology, Sports, Business
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
965
On SlideShare
0
From Embeds
0
Number of Embeds
3
Actions
Shares
0
Downloads
24
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Webinar: Modelling mode and route choices on public transport systems

  1. 1. Modelling Mode and Route Choices on Public Transport Systems Sebastián Raveau Pontificia Universidad Católica de Chile BRT Centre of Excellence Webinar December 5, 2013
  2. 2. Modelling Mode and Route Choices on Public Transport Systems Sebastián Raveau Pontificia Universidad Católica de Chile with the collaboration of: Juan Carlos Muñoz Pontificia Universidad Católica de Chile Juan de Dios Ortúzar Pontificia Universidad Católica de Chile Louis de Grange Universidad Diego Portales Zhan Guo New York University Nigel H.M. Wilson Massachusetts Institute of Technology Carlo Giacomo Prato Technical University of Denmark
  3. 3. It’ is better to use the Yellow Line, but 9 out of 10 use the Red Line! The trip begins by heading in the opposite direction… Destination Origin Attribute Red Line Yellow Line Transfers 1 1 Time 23:40 23:43 Density 5 pax/m2 3 pax/m2 First leg 90 % 50 %
  4. 4. How do we change these travelers’ decision?
  5. 5. Study’s objectives Understanding travellers is essential in Transportation Planning and Design. Identify and quantify the factors that affect the public transport users’ behaviour. Explore differences across modes, in multi-modal public transport networks. Compare the preferences of public transport users in different systems and contexts.
  6. 6. Contents Study Case 1 Metro Networks Study Case 2 Multimodal Network Results & Analysis Extensions & Applications Route Choice Background Conclusions
  7. 7. Route choice modelling Route Choice Background Traditional route choice models usually consider just tangible variables related to the level of service. travel time fare number of transfers These models are sometimes refined including socio-economic variables of the travellers.
  8. 8. Route choice modelling Route Choice Background However, this approach ignores other relevant elements that influence route choice as: comfort and safety transfers accessibility network topology aesthetics These variables are subjective and hard to quantify.
  9. 9. Pathfinding Criteria Route Choice Background
  10. 10. Pathfinding Criteria Route Choice Background
  11. 11. Pathfinding Criteria Route Choice Background Some people follow different criteria when deciding how to get from one point to another: the fastest way the cheapest way avoid walking avoid transferring But most consider many factors at the same time, depending on their preferences and information!
  12. 12. Pathfinding Criteria Route Choice Background
  13. 13. Analyzing travellers decisions on Metro Networks Study Case 1 Metro Networks Santiago London Survey date 2008 1998-2005 Length 78 Km 324 Km Lines 5 11 Stations 85 255 Transfer stations 7 72 Daily trips 2,300,000 3,400,000 Survey size 28,961 16,300
  14. 14. What do people take into account? In-vehicle time Waiting time Walking time (when transferring) Number of transfers Transfer stations layout ascending at level descending Study Case 1 Metro Networks travel time components
  15. 15. Study Case 1 Metro Networks What do people take into account? travel time In-vehicle time Waiting time Walking time (when transferring) Number of transfers Transfer stations layout Transfer stations infrastructure components assisted or semi-assisted or non-assisted and
  16. 16. What do people take into account? In-vehicle time Waiting time Walking time (when transferring) Number of transfers Transfer stations layout Transfer stations infrastructure Mean occupancy Possibility of not boarding Study Case 1 Metro Networks travel time components transfer experience initial occupancy ≥ 75% in London initial occupancy ≥ 85% in Santiago
  17. 17. What do people take into account? In-vehicle time Waiting time Walking time (when transferring) Number of transfers Transfer stations layout Transfer stations infrastructure Mean occupancy Possibility of not boarding Possibility of getting a seat Study Case 1 Metro Networks travel time components transfer experience initial occupancy ≤ 25% in London initial occupancy ≤ 15% in Santiago
  18. 18. What do people take into account? In-vehicle time Waiting time Walking time (when transferring) Number of transfers Transfer stations layout Transfer stations infrastructure Mean occupancy Possibility of not boarding Possibility of getting a seat Route distance Number of stations   Angular cost d  sin    2 Study Case 1 Metro Networks travel time components transfer experience comfort and crowding
  19. 19. Study Case 1 Metro Networks What do people take into account? T2 d2 T1 d1 1 2 d3 Destination Origin  1   d  sin  2  Angular Cost = d1  sin     2 2 2
  20. 20. What do people take into account? Study Case 1 Metro Networks travel time In-vehicle time components Waiting time Walking time (when transferring) Transfer Number of transfers experience Transfer stations layout Easy to obtain! Transfer stations infrastructure comfort and Mean occupancy crowding Possibility of not boarding Possibility of getting a seat Easy to obtain! topological Route distance variables Number of stations Defined based on the schematic maps Angular cost Easy to obtain! Reasonable route
  21. 21. Schematic map’s effect Study Case 1 Metro Networks We want to understand the impact of the Metro network schematic map on the users’ behaviour
  22. 22. Schematic map’s effect Study Case 1 Metro Networks
  23. 23. Set of alternative routes Study Case 1 Metro Networks A key element when dealing with probabilistic route choice models is the definition of the alternatives for the OD pairs of interest Santiago generated based on the actual choices → 2 to 4 alternative routes London generated based on a labeling approach → 2 to 6 alternative routes C-Logit Model for Route Choice
  24. 24. Study Case 1 Metro Networks Estimation results Attribute London Underground Santiago Metro Travel Time - 0.188 - 16.02 - 0.095 - 19.57 Waiting Time - 0.311 - 7.39 - 0.139 - 5.07 Walking Time - 0.216 - 6.14 - 0.155 - 8.23 Number of Transfers - 1.240 - 4.37 - 0.632 - 4.06 Ascending Transfers - 0.138 - 2.57 - 0.323 - 2.73 Even Transfers 0.513 3.53 n. a. (2) n. a. Descending Transfers 0.000 (1) n. a. 0.000 (1) n. a. Assisted Transfers 0.000 (1) n. a. 0.000 (1) n. a.  Semi-Assisted Transfers - 0.328 - 6.83 n. a. (2) n. a. Non-Assisted Transfers - 0.541 - 6.79 - 0.262 - 6.23 Mean Occupancy - 2.911 - 3.48 - 1.018 - 5.60 Getting a Seat 0.098 2.08 0.092 3.41 Not Boarding - 0.430 - 6.06 - 0.380 - 2.97 Angular Cost - 0.065 - 5.87 - 0.024 - 5.48 Map Distance - 0.358 - 5.76 - 0.274 - 5.69 Number of Stations - 0.316 - 5.52 - 0.147 - 3.10 Turning Back - 0.725 - 8.12 - 0.141 - 9.76 Turning Away - 0.968 - 8.00 - 0.226 - 7.11 Commonality Factor - 0.146 - 3.92 - 0.548 - 3.33 Adjusted r 2  0.566 0.382
  25. 25. Marginal rates of substitution Study Case 1 Metro Networks Attribute London Santiago 1 min waiting 1.65 min in-vehicle 1.46 min in-vehicle 1 min walking 1.15 min in-vehicle 1.62 min in-vehicle 1 (basic) transfer 6.60 min in-vehicle 6.63 min in-vehicle 1 % of occupancy 0.16 min in-vehicle 0.11 min in-vehicle Seating 0.52 min in-vehicle 0.97 min in-vehicle Not boarding 2.29 min in-vehicle 3.99 min in-vehicle 1 station 1.68 min in-vehicle 1.54 min in-vehicle Turning back 3.86 min in-vehicle 1.48 min in-vehicle Turning away 5.15 min in-vehicle 2.37 min in-vehicle
  26. 26. Study Case 1 Metro Networks Marginal rates of substitution Transfer valuations in London Getting a seat Intermediate Not boarding Assisted 06.81 min 07.33 min 09.62 min Semi-assisted 08.56 min 09.07 min 11.36 min Non-assisted 09.69 min 10.21 min 12.49 min 03.35 min 03.87 min 06.15 min Assisted 06.08 min 06.60 min 08.88 min Semi-assisted 07.82 min 08.34 min 10.63 min Non-assisted 08.95 min 09.47 min 11.76 min Transfer Type Ascending At level Descending
  27. 27. Study Case 1 Metro Networks Marginal rates of substitution Transfer valuations in Santiago Getting a seat Intermediate Not boarding Assisted 09.05 min 10.02 min 14.01 min Non-assisted 11.80 min 12.77 min 16.76 min Assisted 05.67 min 06.63 min 10.62 min Non-assisted 08.41 min 09.38 min 13.37 min Transfer Type Ascending Descending range in London 3.35 to 12.49 min range in Santiago 5.67 to 16.76 min
  28. 28. Transantiago - Santiago, Chile Study Case 2 Multimodal Network 34 communes 7 million people 700 sq Km 10 million daily trips 55% in public modes
  29. 29. Study Case 2 Multimodal Network Transantiago - Santiago, Chile 10 zones feeder bus lines trunk bus lines express bus lines Metro
  30. 30. Transantiago - Santiago, Chile Study Case 2 Multimodal Network 30,000 daily trips (7am to 12 pm) 1% of all the city trips 1,892 respondents access to all modes
  31. 31. Analyzing travellers decisions on Transantiago Study Case 2 Multimodal Network The objective is to expand the behavioural models obtained form Metro, to the entire public transport system. Some new explanatory variables are: fare distinguish travel time by mode distinguish transfers by modes involved variability of in-vehicle and waiting times When travelling in frequency-based networks, the travellers might follow different route choice strategies.
  32. 32. Study Case 2 Multimodal Network Route choice strategies Choosing a itinerary Choosing an hyper-path → considering common lines
  33. 33. Route choice strategies Study Case 2 Multimodal Network We found that 66.6% of the travellers that could choose their routes considering common lines, didn’t do so... One might argue that considering common lines is a personal characteristic, rather than the behaviour of everyone. We propose modelling two types of individuals: Those who consider common lines Those who don’t consider common lines
  34. 34. Study Case 2 Multimodal Network Logit probability of considering common lines Attribute Parameter t-Value Income – More than 1,000€/month - 0.940 3.22 Income – 500€/month to 1,000€/month - 0.327 3.45 Income – Less than 500€/month - 0.000 base Frequency - Al least once a week - 1.322 4.98 Frequency - Al least once a month - 0.766 3.71 Frequency – Rarely/Never - 0.000 base Age – Less than 30 years old - 0.399 2.90 Age – More than 30 years old - 0.000 base Constant - 2.051 - 5.76 Log-Likelihood - 800.66 r2 0.525
  35. 35. Study Case 2 Multimodal Network Mode/route choice results Consider Common Lines Variable Fare (CLP) In-vehicle time (min) Waiting time (min) Walking time (min) Bus-bus transfer Bus-Metro transfer Metro-Metro transfer Travelling seated Not boarding Log-Likelihood r2 Parameter - 0.041 - 0.625 - 1.601 - 1.856 - 2.822 - 2.201 - 1.939 1.886 - 1.890 Do Not Consider Common Lines t-value Parameter - 2.32 - 0.050 - 2.17 - 0.477 - 4.37 - 1.217 - 2.11 - 1.353 - 2.98 - 2.139 - 2.32 - 1.849 - 2.33 - 1.673 2.88 1.652 - 1.97 - 1.533 - 1,512 0.487 t-value - 2.45 - 2.39 - 3.78 - 2.43 - 2.23 - 2.63 - 2.09 2.33 - 2.04
  36. 36. Marginal rates of substitution Study Case 2 Multimodal Network Variable Consider Common Lines Do Not Consider Common Lines In-vehicle time (min) Waiting time (min) Walking time (min) Bus-bus transfer Bus-Metro transfer Metro-Metro transfer Travelling seated Not boarding € 1.35 per hour € 3.51 per hour € 4.06 per hour € 0.11 per transfer € 0.08 per transfer € 0.07 per transfer € 0.07 per leg € 0.07 per vehicle € 0.88 per hour € 2.25 per hour € 2.50 per hour € 0.07 per transfer € 0.06 per transfer € 0.05 per transfer € 0.05 per transfer € 0.05 per transfer Those who consider common lines are more sensitive to the different attributes.
  37. 37. Using the model for policy Change in the Santiago Metro Map Extensions & Applications
  38. 38. Some extensions to this work Apply the model to different cities and systems Extensions & Applications
  39. 39. Some extensions to this work Map design optimization Extensions & Applications
  40. 40. Some extensions to this work Application to journey planner Extensions & Applications
  41. 41. What did we learn today? Conclusions Public transport users take into account a wide variety of attributes when choosing routes. The modelling effort should be on what we can explain, rather than in what we can’t explain. Network’s topology, and specially the way it’s presented to users on a daily basis, is relevant. Different individuals follow different strategies when choosing routes.
  42. 42. What did we learn today? Conclusions Don’t forget that we are dealing with individuals, whose behaviour is hard to understand and model
  43. 43. Modelling Mode and Route Choices on Public Transport Systems Sebastián Raveau Pontificia Universidad Católica de Chile BRT Centre of Excellence Webinar December 5, 2013

×