LT5: Modeling reliability, cost, travel times, safety, comfort and other relevant variables of modal choice

1
LT5 Modeling reliability, cost, travel times,
safety, comfort and other relevant variables of
modal choice  

• Juan Carlos Muñoz, Juan de Dios Ortúzar and
  Sebastián Raveau
• Departamento de Ingeniería de Transporte y
  Logística
• Pontificia Universidad Católica de Chile

2
How do transit users choose their routes a case
study of Metro in Santiago

• Sebastián Raveau Juan Carlos Muñoz Louis de
  Grange

3
Only 11 travel through Santa Ana!!
The traveler starts heading in the opposite
direction
The map influences the Baquedano trip
Destination
Origin
Attribute Baquedano Santa Ana
Travel Time 2340 2343
Density 5 pax/m2 3 pax/m2
Occupancy 88 49
Transfers 1 1
4
Contents
Results
Study Case
Background
Explanatory Variables
Applications
Extensions
5
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 travelers

6
Background

• However, this approach ignores other relevant
  elements that influence route choice as
• comfort
• safety
• transfers quality
• network topology
• These variables are subjective and hard to
  quantify

7
Models Variables

• Travel time
• Waiting time
• Number of transfers
• Walking time when transferring
• Ascending level transfers
• Transfers without escalators
• Mean occupancy rate
• Network knowledge
• Possibility of not boarding
• Possibility of getting a seat
• How direct is the route (Angular cost)

traditional variables
physical characteristics
(volume as proxy)
(initial occupancy 85)
(initial occupancy 15)
8
Models Variables

• What should the angular cost satisfy
• minimum for 0º and maximum for180º
• small marginal variations at these extremes
• (non-linear effect)
• grow with the distance covered
• A specification that satisfies these conditions
  is

9
Models Variables
T2
T1
Destination
Origin
10
Models Variables

• Travel time
• Waiting time
• Number of transfers
• Walking time when transferring
• Ascending level transfers
• Transfers without escalators
• Mean occupancy rate
• Network knowledge
• Possibility of not boarding
• Possibility of getting a seat
• Angular cost
• Turning back to the origin
• Turning away from the destination

traditional variables
physical characteristics
easy to obtain!
network topology
easy to obtain!
11
Study Case

• Route choice in the Santiago Metro network
• 5 lines and 85 stations
• 2,300,000 daily trips
• Period morning peak 700 900 hrs
• evening peak 1800 2000 hrs
• 790,000 daily trips in peak hours

12
Study Case

• Survey conducted by Metro on October, 2008
• Total respondents 92,800
• Users that transfer 42,700 (44 )
• One unique route 26,900
• Two or more routes 15,800 (37 )

13
Study Case

• We also want to understand the impact of the
  Metro network schematic map on the users behavior

14
Results

• Multinomial Logit Model for the route choice
• For every OD pair, the choice set was given by
  routes traveled by at least one person
• We will compare two models
• Base Model Proposed Model
• traditional variables ? ?
• physical characteristics ? ?
• network topology ? ?

15
Results
Attribute Base Model Base Model Proposed Model Proposed Model
Travel time - 0.144 - 46.61 - 0.117 - 27.29
Waiting time - 0.203 - 3.68 - 0.121 - 3.17
Walking time - - - 0.229 - 6.87
Number of transfers - 1.220 - 17.06 - 0.420 - 3.69
Ascending transfers - - - 0.432 - 8.02
Transfers without escalator - - - 0.457 - 12.34
Occupancy rate - - - 1.250 - 3.05
Network knowledge - - 0.030 6.81
Possibility of not boarding - - - 0.431 - 6.06
Possibility of getting a seat - - 0.106 1.81
Angular cost - - - 0.038 - 6.75
Turn back to the origin - - - 0.503 - 10.65
Turn away from destination - - - 0.512 - 11.59
Log-Likelihood - 7,416 - 7,416 - 7,136 - 7,136
16
Results
Attribute Real Distances Real Distances Map Distances Map Distances
Travel time - 0.119 - 27.94 - 0.117 - 27.29
Waiting time - 0.111 - 3.17 - 0.121 - 3.17
Walking time - 0.240 - 7.14 - 0.229 - 6.87
Number of transfers - 0.449 - 3.95 - 0.420 - 3.69
Ascending transfers - 0.426 - 7.89 - 0.432 - 8.02
Transfers without escalator - 0.438 - 11.95 - 0.457 - 12.34
Occupancy rate - 1.430 - 3.53 - 1.250 - 3.05
Network knowledge 0.031 7.14 0.030 6.81
Possibility of not boarding - 0.426 - 6.00 - 0.431 - 6.06
Possibility of getting a seat 0.126 2.16 0.106 1.81
Angular cost - 0.024 - 5.78 - 0.038 - 6.75
Turn back to the origin - 0.516 - 10.96 - 0.503 - 10.65
Turn away from destination - 0.505 - 11.43 - 0.512 - 11.59
Log-Likelihood - 7,142 - 7,142 - 7,136 - 7,136
17
Results

• Marginal rates of substitution
• There is a bias when relevant variables are not
  included

Variable Base Model Proposed Model
1 min of wait 1.41 min of travel 1.04 min of travel
1 min of walk n. a. 1.96 min of travel
18
Results

• Marginal rates of substitution
• Base Model 8,1 min of travel

Transfer Type Transfer Type WithEscalator Without Escalator
Possibility of Sitting Descending 2.7 min of travel 6.6 min of travel
Possibility of Sitting Ascending 6.4 min of travel 10.3 min of travel
Intermediate Descending 3.6 min of travel 7.5 min of travel
Intermediate Ascending 7.3 min of travel 11.2 min of travel
Possibility of not Boarding Descending 7.3 min of travel 11.2 min of travel
Possibility of not Boarding Ascending 11.0 min of travel 14.9 min of travel
19
Applications
Predict route choice for all trips within a set
of origins and a set of destinations
7,300 trips (peak)
route 1 Baquedano route 2 Tobalaba route 3
Santa Ana route 4 La Cisterna
20
Applications

• Assignment results
• MSE Base Model 18.9
• MSE Proposed Model 10.2

Assignment Baquedano Santa Ana La Cisterna Tobalaba
Observed Trips 5,362 99 30 1,854
Mínimum Time 6,676 668 0 0
Base Model 3,884 361 33 3,067
Proposed Model 4,446 270 25 2,603
21
Extensions

• Compare the results and forecasting with other
  models used in Transport Systems Planning
• Application to a more dense network
• Base Proposed
• MSE OD pairs with 2 alternatives 7,9
  6,7
• MSE OD pairs with 3 alternatives 20,0
  10,7
• MSE OD pairs with 4 alternatives 27,8
  15,9

22
Extensions

• Compare the results and forecasting with other
  models used in Transport Systems Planning
• Application to a more dense network
• Application to a more distorted network
• correlation of distances Santiago 94
• correlation of distances London 22

23
Extensions
24
Extensions
Attribute Santiago Model Santiago Model London Model London Model
Travel time - 0.106 - 29.82 - 0.158 - 18.16
Waiting time - 0.117 - 3.30 - 0.132 - 11.28
Walking time - 0.210 - 6.79 - 0.155 - 11.45
Number of transfers - 0.663 - 8.17 - 0.463 - 5.59
Ascending transfers - 0.335 - 6.97 0.108 2.40
Transfers at level - - 0.232 5.09
Transfers without escalator - 0.409 - 12.67 0.211 4.79
Angular cost - 0.041 - 7.57 - 0.020 - 0.56
Turn back to the origin - 0.529 - 11.45 - 0.354 - 6.29
Turn away from destination - 0.576 - 13.55 - 0.467 - 6.522
Sample Size 16,029 16,029 2,721 2,721
The absence of flow-related variables
bias the results What other factors can affect
the choice of transfer stations?
25
Extensions

• Compare the results and forecasting with other
  models used in Transport Systems Planning
• Application to a more dense network
• Application to a more distorted map
• Map design optimization

26
Extensions
27
How do transit users choose their routes a case
study of Metro in Santiago

• Sebastián Raveau Juan Carlos Muñoz Louis de
  Grange

28
Publications and working papers

• Raveau, S., J.C. Muñoz, and L. de Grange (2011) A
  topological route choice model for metro.
  Transportation Research Part A, Vol 45 (2),
  138147
• Raveau, S., Z. Guo, J.C. Muñoz and N.H. Wilson.
  (2012) Route Choice Modelling on Metro Networks
  time, Transfer, crowding, and topology. To be
  submitted to Transportation Research Part A.
• Navarrete, F. and J. de D. Ortúzar (2012)
  Subjective valuation of the transit transfer
  experience the case of Santiago de Chile.
  Submitted to Transport Policy.

29
Conferences and seminars

• Raveau, S., J.C. Muñoz y J. de D. Ortúzar (2012)
  Modelling Mode and Route Choices on Public
  Transport Systems. Submitted to the International
  Symposium of Traffic Theory and Transportation to
  be held in the Netherlands in 2013.
• Raveau, S., J.C. Muñoz y L. de Grange (2011)
  Modelación y análisis temporal de elección de
  ruta en Metro. XV Congreso Chileno de Ingeniería
  de Transporte, Santiago, Chile.
• Raveau, S., Muñoz, J.C. and de Grange, L. (2011)
  A topological route choice model for metro.
  Transport for London, London, UK.
• Raveau, S., Muñoz, J.C. y de Grange, L. (2010) El
  efecto de la topología de la red y las
  percepciones en la elección de ruta. XVI Congreso
  Panamericano de Ingeniería de Tránsito y
  Transporte. Lisboa, Portugal.

30
2012
31
In-progress or future research I

• Comparison of route choice models for Metro of
  London and Santiago. A paper should be submitted
  to Trans Res A this month.
• Extend the study for a route and mode choice
  models within a transit system. We made an 1,800
  people survey in Santiago with this purpose. This
  research is being developed by PhD student
  Sebastian Raveau.
• Develop a similar survey in Bogota, Colombia to
  understand the role played by a BRT-based network
  in passengers choices.
• Compare results between Santiago and Bogota to
  understand how much of a Metro service BRT
  provides in Bogota.

32
In-progress or future research II

• Build a tactic tool to predict passenger flows in
  a multimodal transit system. Such a model would
  have to deal with endogeneity since passengers
  flows affect travelers choices.
• Build a tool to advise passengers how to travel
  in a complex multimodal transit system.
• Develop a methodology to feed our route and mode
  choice models with feedback provided by users of
  the Passenger Travel Advising Tool.

33
Grants obtained

• FONDEF (2012-2014). A tactic-strategic tool for
  urban transit systems planning and management.
  Total funding of US800,000. Involvement of Metro
  and Alsacia (bus operator)
• Juan Carlos Muñoz and Juan de Dios Ortúzar
• PUC (2011-2012). Interdisciplinary research
  project to understand how to improve the users
  transfer experience on a transit system
  (US10,000).
• Ricardo Giesen, Juan de Dios Ortúzar, Juan
  Carlos Muñoz, Patricia Galilea, Juan Carlos
  Herrera, Margarita Greene, Rossana Forray, José
  Allard

34
From the papers to the streets

• Several interviews with the media during 2011
• Interviews with Metro and government authorities
  during 2011
• Metro de Santiago changed its map based on our
  results to induce a more socially optimal
  behavior

35
Applications

• Changes in the Santiago Metro Map

36
From the papers to the streets

• Several interviews with the media during 2011
• Interviews with Metro and government authorities
  during 2011
• Metro de Santiago changed its map based on our
  results to induce a more socially optimal
  behavior
• We are now working un using our assignment model
  to design interchange stations and predict flows
  for the Metro network in Santiago that will
  consider two extra lines in 2016.