1. Using Transit ITS Data for Service Planning
University of Arizona, Transit Research Unit
Alireza Khani, Sanggu Lee, Hyunsoo Noh, Neema Nassir, Mark Hickman
ITS Arizona 18th Annual Conference, Sep 29, 2011

2. Outline
Introduction / Motivations
Data
Data Analysis
Applications

3. Introduction / Motivation
How to transform the massive data into useful information and support decision-making
From fare card data,
how to identify boarding stops
how to infer alighting stops (known boarding)
how to estimate travel/alighting time
how to analyze travel pattern/behavior
From Google’s transit data,
how to utilize a transit network
how to develop an intermodal network
how to find the shortest/optimal path

4. Automated Data Collection System Data
Metro Transit ( data
Serving Minneapolis/St. Paul area, MN
November 2008 (30 days)
Entry-only-control operation
Implemented independently
Automatic Vehicle Location (AVL)
Automatic Passenger Counting (APC)
Automatic Fare Collection (AFC)
Bus location based on GPS
Bus systems based on sensors in doors
Contactless smart card with unique ID
Information (location, ..) of
a given vehicle
Passenger boarding/alighting counts for stops
Entry information (time, route, …)
at the individual level
4.2 M Records
3.4 M Records
2.2M Records

5. Google’s Transit Data General Transit Feed Specification (GTFS)
Open to public and frequently updated by agencies
Stops, routes, trips, stop_times (schedules)
Metro Transit in 2008
Routes.txt: 220
Stops.txt: 14,601
Trips.txt: 25,417
Stop_times.txt: 1,471,150

6. Automated Data Collection System (ADCS)
Relational Schema
Automated Data Collection System (ADCS)
STOP TIMES
CALENDAR
TRIPS
ROUTES
STOPS
TRAVELER
ROUTE NO
DATE / TIME
USE TYPE
VEHICLE ID
LOCATION
AFC
ROUTE ID
ARRIVAL TIME
TRIP ID
DEPARTURE TIME
STOP ID
SEQUENCE
APC with LOCATION
SCHEDULE
STOP LOCATION
STOP CITY
DESCRIPTION
SERVICE ID
GENERAL TRANSIT FEED SPECIFICATION (GOOGLE)
VEHICLE
Presented at 90th TRB Annual Meeting in Washington, DC on January 2011

7. Presented at 90th TRB Annual Meeting in Washington, DC on January 2011
Data Analysis
DURATION 1
DURATION 2
DURATION 3
TRANSACTION 1
DATE / TIME
ROUTE NUMBER
USE TYPE
BUS ID
LOCATION
USER
SPECIAL SERIAL NO
Purchases
FARE CARD
CARD TYPE
METRO PASS
U-PASS
C-PASS
Requests
STORED VALUE
TRANSACTION 2
TRANSACTION 3
TRANSACTION 4
Access Time
Waiting Time
Activity Time
Transfer Time
Egress Time
In-Vehicle Time
Presented at 90th TRB Annual Meeting in Washington, DC on January 2011

8. Data Analysis – cont.
Between-day Hourly Variation in Transactions by Card Type
Duration Distribution by Card Type
Presented at 90th TRB Annual Meeting in Washington, DC on January 2011

9. Applications Application Conference Name Travel Pattern Analysis
90th TRB Annual Meeting
Transit O-D Estimation
90th TRB Annual Meeting (TRR in press)
Transit Path Choice Model
13th TRB National Planning Applications
Transit Path Algorithms
Trip-Based Shortest Path
91st TRB Annual Meeting (under review)
Intermodal Optimal Path
Intermodal Tour Path
Link-Based Transit Hyperpath
Transit Assignment and Simulation
Integration of Land Use and Transportation
Temporal and Spatial Linkage between
Land Use and Transit Demand
1st World Symposium for Transport and Land Use Research
Stop Aggregation Model

10. Presented at 90th TRB Annual Meeting in Washington, DC on January 2011
Application 1
Transit O-D Estimation Using Fare Card
Objective: To find the origin, destination, and transfer stops of each passenger in Minneapolis/St. Paul transit system
Bus
Walk
Transaction
Bus Stop
Home
Shopping
Work
Transfer
Presented at 90th TRB Annual Meeting in Washington, DC on January 2011

11. Presented at 90th TRB Annual Meeting in Washington, DC on January 2011
Application 1
Transit O-D Estimation – cont.
This model
Finds the boarding and alighting stop for each AFC transaction, as well as the possible trip taken by the passenger
Infers the transfers made by passenger by looking at the boarding and alighting times, service headway, etc
Verifies the results by comparing with APC sample data
Results for a typical weekday
90,154 transactions in AFC data set
84,413 transaction after refining the data set
33,514 transactions with estimated OD (28,260 persons)
98% percent verified by APC sample data
Presented at 90th TRB Annual Meeting in Washington, DC on January 2011

12. Presented at 90th TRB Annual Meeting in Washington, DC on January 2011
Application 1
Transit O-D Estimation – cont. AM
Mid-day PM
Origin
Destination
Presented at 90th TRB Annual Meeting in Washington, DC on January 2011

13. Transit Path Choice Model Using Fare Card
Application 2
Transit Path Choice Model Using Fare Card
Objective: To calibrate a logit model for transit path choice using the results of the O-D estimation study (Minneapolis/St. Paul)
Origin
Destination
Choice set
Origin
Destination
Passenger 1
Origin
Destination
Passenger 2
Presented at 13th TRB National Planning Applications Conference in Reno, NV on May 2011

14. Transit Path Choice Model – cont.
Application 2
Transit Path Choice Model – cont.
Attribute
Definition
In Vehicle Time VT
Sum of the times spent on rides of all legs of the path
Number of Transfers TR
Number of bus transfers for the path
Waiting Time WT
Sum of waiting times for all the transfers in the path
Walking Distance WD
Sum of walking distances for all the transfers in the path
Express Route EX
Indicates whether path contains any express routes or not
Downtown Route DT
Indicates whether path contains a leg in downtown or not
Covers Express
CEX
Indicates whether the user’s pass covers the express fare or not
Covers Downtown
CDT
Indicates whether the user’s pass covers the downtown fare or not
Presented at 13th TRB National Planning Applications Conference in Reno, NV on May 2011

15. Transit Path Choice Model – cont.
Application 2
Transit Path Choice Model – cont.
Calibration Tools
Easy Logit Modeler (ELM) (
Biogeme (
Results
Time Period
# of Records
Model
Rho2
t-statistics
Rush-Hours
1225 TR
0.029
-6.41
Non-Rush Hours
695 TR
0.038
-4.76
All-Day
1922 TR
0.032
-7.96
Presented at 13th TRB National Planning Applications Conference in Reno, NV on May 2011

16. Transit Path Algorithms Using GTFS Data
Objective: To develop efficient algorithm for transit and intermodal shortest/optimal path
Trip-Based Shortest Path (TBSP)
Shortest path algorithm utilizing hierarchical transit network
Intermodal Optimal Path (IOP)
Optimal path with combined modes (auto and transit)
Intermodal Optimal Tour (IOT)
Optimal path for a tour with multiple destinations with combined modes
Link-Based Transit Hyperpath
Optimal strategy for schedule-based transit systems

17. Improvement in Label Setting Improvement in Label Correcting
Application 3
Trip-Based Shortest Path (TBSP)
The idea
Using trips as the elements of the network (instead of links)
Distinguish between transfer stops and non-transfer stops in the labeling algorithm
Performance in computation
Sacramento, CA
San Francisco (MUNI)
No. of Stops (transfer)
2880
(1028)
4424
(2865)
Improvement in Label Setting
83%
54%
Improvement in Label Correcting
32%
26%

18. Intermodal Optimal Path (IOP)
Application 4
Intermodal Optimal Path (IOP)
This model
Using TBSP for the transit side
Using a multi-source time-dependent shortest path for the auto side
Modeling Park-and-Ride locations to make the connections
Contributions
Solve for the optimal intermodal path and transfer point between an origin and a destination for a preferred arrival time
The complexity is the same as single-mode path (e.g. TBSP)
Destination
Origin

19. Intermodal Optimal Tour (IOT)
Application 5
Intermodal Optimal Tour (IOT)
Objective
Capture the effect of back trip on park-and-ride choice
Find optimal tour and optimal park-and-ride facilities for a sequence of activities
Origin D1 D2 P1 P2 D3
D10
D20
P10
P20
D11
D12
D22
D21
P11
P22
D32
D31
D30

20. Link-Based Transit Hyperpath
Application 6
Link-Based Transit Hyperpath
Objective: To search an optimal strategy path on a transit schedule network using logit model
Link-Based Time-Expanded Network
A run (or trip) segment between two consecutive stops is considered a unique link
Search Model e1
[cost, departure time]
[25, 7] e5
[6, 8] e6
[6, 19]
e10
[4, 15] e2 es
[7, 1] e7
[6, 28] et
e11 A B C
[4, 21] D e3
[8, 5] e8
[4, 11]
e12
[10, 13] e9
e13 e4
[5, 16]
[10, 23]
[7, 10]
e14
[10, 43]
Route 1
Route 3
Route 2
Route 4

21. Transit Assignment and Simulation
Application 7
Transit Assignment and Simulation
Integration among transportation models
Activity-Based Model (ABM) for demand
Dynamic Traffic Assignment (DTA) model
FAST-TrIPs
Passenger assignment and simulation on GTFS
Transit Assignment and Simulation
ABM
DaySim OpenAMOS
DynusT MALTA
Vehicle assignment and simulation
DTA
Presented at 13th TRB National Planning Applications Conference in Reno, NV on May 2011

22. Transit Assignment and Simulation – cont.
Application 7
Transit Assignment and Simulation – cont.
Contributions
Capability to model travel behaviors according to Google’s GTFS which allows to connect to DTA model, such as boarding and alighting
As well as integrating between ABM and DTA, the schedule-based transit assignment and simulation supports the intermodal capability
Funded by FHWA EAR program and SHRP2 C-10B

23. Thank you!!
University of Arizona, Transit Research Unit