Open data in the General Transit Feed Specification (GTFS) format has led to many innovations in the transit industry. One of these innovations has been the emergence of open-source software projects that utilize open transit data and offer various multi-modal traveler information services. OneBusAway (http://onebusaway.org/) started as a student project at the University of Washington, and now offers real-time transit arrival information riders at more than 10 cities around the world. OpenTripPlanner (http://www.opentripplanner.org/) started as a project in TriMet, OR and has been used for the basis of many other trip planning applications world-wide, including the university campus-centric USF Maps App (http://maps.usf.edu/). This presentation will discuss the evolution and benefits of the OneBusAway and USF Maps App, including the ability for anyone to deploy these projects in new locations.
1. Center for Urban Transportation Research | University of South Florida
TDM Technology Session
Sean J. Barbeau, Ph.D.
Principal Mobile Software Architect for R&D
Center for Urban Transportation Research
University of South Florida
National Center for Transit Research
2. 2
Agenda
• OneBusAway – How does real-time
information affect riders?
– Slide credits to Dr. Kari Watkins, Georgia Tech
• USF Maps App – Multimodal campus-focused
solution
4. 4
What is OneBusAway?
• What? Suite of tools that provides real-
time bus/train tracking information
– Open source software
– API for developers
– Free to riders
• Why? Make riding public transit easier by
providing good information in usable
formats
– Research to evaluate the impacts
4
5. 5
Mobile Apps!
Android Windows PhoneiPhone
Support user location, route, stop contextual /personalized information
All OPEN-SOURCE!
7. 7
Seattle, WA:
Original deployment
New York, NY:
Adapted for the MTA
(Bus Time)
Washington, DC:
2016
Atlanta, GA:
2013
Tampa, FL:
2013
York, ON:
2014
Rouge
Valley, OR:
2015
Where is OneBusAway?
San Joaquin, CA:
In testing
San Diego, CA:
2016
Lappeenranta,
Finland:
In testing
9. 9
Impacts
• Riders are more satisfied
• Riders feel safer
• Riders wait less time
• Do they take more transit trips?
10. 10
Change in Satisfaction
“I no longer sit with
pitted stomach
wondering where
is the bus. It's less
stressful simply
knowing it's nine
minutes away, or
whatever the
case.”
11. 11
Perception of Safety
• Perception of Safety
– 79% no change
– 18% somewhat safer
– 3% much safer
• Safety correlated with
gender
– χ2=19.458
– p-value=0.001
0% 20% 40% 60% 80% 100%
Men
Women
Somewhat Less Safe
No Change
Somewhat More Safe
Much Safer
11
12. 12
Wait Time
• Without real time, perceived wait > actual wait
• With real time, perceived wait = actual wait
• Value of real time >> more frequent service
Group Real Time Schedule Difference T-stat
(p-value)
Mean Typical Wait 7.54 9.86 2.32 5.50 (0.00)
Aggravation Level 3.35 3.29 -0.05 -0.24 (0.81)
Actual Wait Time 9.23 11.21 1.98 2.17 (0.03)
12
13. 13
Ridership - Tampa
Before-After Control Group
Research Design
• Motivation: HART provided USF & Georgia
Tech special access to real-time data
• Recruitment: HART website/email list
(Incentive of 1 day bus pass)
• Measurement: Web-based surveys
• Group Assignment: Random number
generator
• Treatment: OneBusAway
Limiting the Treatment: iPhone
& Android Apps
14. 14
Tampa
• Significant improvements in the waiting experience
– Decreases in self-reported usual wait times
– Increases in satisfaction with wait times and reliability
• Little evidence supporting a change in transit trips
– Approx. 1/3 of RTI users stated they ride the bus more frequently, perhaps because
of:
• Affirmation bias of respondents
• Scale of measurement (trips per week)
– Only riders within sphere of transit agency
15. 15
Ridership - New York City
#1. February 2011:
Brooklyn Pilot (B63)
#2. February 2012:
Staten Island Launch
#3. November 2012:
Bronx Launch
#4. October 2013:
Manhattan Launch
#5. March 2014:
Queens + Brooklyn Launch
16. 16
Ridership - New York City
• Method
• Comparison of multiple panel regression techniques in a well-suited natural
experiment
• Conclusions
Real-time Information as a single variable
• Average increase of ~115 rides per route per weekday (median of 1.6%), similar to
previous Chicago study
Real-time Information by route size
• Average increase of ~338 rides per weekday on the largest quartile of routes
(median of 2.3%)
• Limitations
• Short Timescale
• Aggregate Analysis
17. 17
Comparison of Key Findings
New York City Tampa Atlanta
Transit Agency
Methodology
Natural experiment with
panel regression
Behavioral experiment with a
before-after control group design
Before-after analysis of transit
trips
Key Finding
Average weekday route-
level increase of ~115 rides
(median of 1.6%);
Average weekday increase
of ~338 rides on the largest
routes (median of 2.3%)
Little evidence supporting a change
in bus trips;
Significant improvements in the
waiting experience, particularly wait
times
Little evidence supporting a
change in bus/train trips;
Perceived improvements in
wait times and overall
satisfaction with MARTA
19. 19
Background
• USF students have many travel options:
– Drive
– USF Bull Runner
– Hillsborough Area Regional Transit
– Bike
– Share-A-Bull Bike share
– Walk
• For those unfamiliar with campus (and even those
that are), the best option for each trip isn’t obvious
20. 20
Background (Con’t)
• Transit and bike share modes also have a real-
time component
• Knowing where USF buildings are, and how to
get from A to B, is challenging
– Requires translating 3 letter abbreviation into
building name and location
• How can we make getting around USF campus
easier for students, staff, and visitors?
21. 21
USF Student Green Energy Fund
(SGEF)
• Initially funded two student-driven projects:
– Smart Parking
– “Share-A-Bull” Bike share
• USF Maps App was created to share
information on all modes with
students/staff/visitors
• Funding from FDOT to supervise students
36. 36
Other features
• Walking paths that avoid stairs
– Useful for those with limited mobility (e.g., in
wheelchairs)
• Bike paths that prefer bike lanes
• Transfer from Bull Runner to HART (and PSTA) buses
– Students ride free on HART
• All open-source software
– Based on OpenTripPlanner.org
– Can continue to add new features
• Can deploy at multiple university sites
– e.g., Different USF campuses, small communities
37. 37
Open data powers these apps
• OneBusAway
– General Transit Feed Specification (GTFS)
– GTFS-realtime
• USF Maps App
– GTFS
– GTFS-realtime
– General Bikeshare Feed Specification (GBFS)
– OpenStreetMap data
38. 38
Set up your own version!
• Requires some technical expertise
– Experience in setting up servers (Tomcat) a plus
– If you want to modify things, experience with
Java/Javascript is very useful
• Most IT departments should have the required
skillset to get a demo up and running
39. 39
Set up your own OneBusAway!
• You’ll need:
– GTFS data
– If you want real-time, one of the following:
• GTFS-realtime TripUpdates feed
• SIRI
• Other formats - http://bit.ly/OBARealtimeFormats
• Instructions - http://bit.ly/OBAQuickStart
40. 40
Set up your own USF Maps App!
• You’ll need:
– GTFS data for planning transit trips
– If you want real-time bus locations:
• GTFS-realtime VehiclePositions feed
– If you want bikeshare locations/trip planning:
• GBFS data
– Walking/bike paths:
• OpenStreetMap data
– If you want Layers:
• OpenStreetMap data
– Bike lanes, bike repair, parking lots, vehicle charging stations
• Car share – update an XML file
• Emergency phone locations - a config file with locations
– Building abbreviations
• Update an XML file with abbreviations/locations
– Instructions - http://bit.ly/USFMapsInstructions
41. 41
Thanks!
Sean J. Barbeau, Ph.D.
barbeau@cutr.usf.edu
813.974.7208
OneBusAway partners = Dr. Kari Watkins (GA Tech), Dr. Candace Brakewood (CCNY), Dr. Brian Ferris, Dr.
Alan Borning (UW), Sound Transit, KC Metro, Pierce Transit, MTA NYC, HART, PSTA, MARTA, ARC,
independent developers, many more…
OneBusAway funding = NSF, NCTR, US DOT, NCTSPM, CUTR, GVU Center, IPAT, and more…
Current USF Maps App Developers – Joseph Fields and JB Subils
USF Maps App funding partners - USF Student Green Energy fund and Florida Department of
Transportation
42. 42
References
• Ferris, Brian, Kari Watkins, and Alan Borning. “OneBusAway: Results from providing real-time arrival information for public transit.”
Proceedings of Association for Computing Machinery Conference on Human Factors in Computing Systems (CHI) 2010.
• Watkins, Kari, Brian Ferris, Alan Borning, G. Scott Rutherford and David Layton. “Where Is My Bus? Impact of mobile real-time
information on the perceived and actual wait time of transit riders.” Transportation Research Part A, Vol. 45, No. 8, 2011.
• Gooze, Aaron, Kari Watkins and Alan Borning. “Benefits of Real-Time Transit Information and Impacts of Data Accuracy on Rider
Experience”, Transportation Research Record #2351, 2013.
• Windmiller, Sarah, Todd Hennessy and Kari Watkins, “Accessibility of Communication Technology and the Rider Experience: Case Study
of St. Louis Metro” Transportation Research Record #2415, 2014.
• Barbeau, Sean, Alan Borning and Kari Watkins, “OneBusAway Multi-region – Rapidly Expanding Mobile Transit Apps to New Cities”
Journal of Public Transportation, Vol. 17, No. 4, 2014.
• Brakewood, Candace, Sean Barbeau and Kari Watkins, “An experiment validating the impacts of transit information on bus riders in
Tampa, Florida”, Transportation Research Part A, Vol. 69, 2014
• Brakewood, Candace, Gregory Macfarlane, and Kari Watkins, “The Impact of Real-time Information on Bus Ridership in New York City”,
Transportation Research Part C, Vol. 53, 2015.
• Berrebi, S., K. Watkins, and J. Laval, “A Real-Time Bus Dispatching Policy to Minimize Headway Variance”, Transportation Research Part
B, Vol. 81, pp. 377-389, 2015.
Editor's Notes
NCTR is a Tier 1 University Transportation Research Center at USF
Evidence supporting changes in the number of transit trips associated with real-time information is limited.
First, it is possible that the “revealed” questions suffered from measurement error and did not capture sufficient levels of detail. For example, the use of trips per week to measure transit travel frequency could be insufficient if a person only makes one or two additional trips per month attributable to RTI. A more reliable way to measure this would be to record trips over an extended period of time (e.g. have respondents report their number of trips per week for all the weeks over the study period). Similarly, a five-point Likert scale for satisfaction may not be sufficiently refined to capture a small increase in satisfaction associated with using RTI.
A second plausible explanation is bias on behalf of the survey respondents when answering the stated questions. The survey methods literature has shown that respondents often have a social desirability bias and will provide an affirmative response that may not align with their actual behavior
We found that >90% were more satisfied with transit even though nothing else was done except to give them more information.
One of the most interesting aspects was perception of safety. Although only 20% felt safer overall, there was a correlation with gender. 30% of women felt safer using the bus as a result of having real time information.
The result was that when people did not have real time information, they perceived that they waited longer than they actually did. But with real time information, there was no longer a statistical difference in wait time. In fact, in a regression analysis, the value of real time information was greater than more frequent service until service was every 10 minutes or less. The typical wait time was found to be less with real time information, but the aggravation level was not (counter to our expectation). This could be because people that seek out real time have a higher level of aggravation in the first place. The biggest finding was that the actual wait time was less as well, meaning that people did not even arrive at the stop until closer to their buses arrival if they had real time information to know when it was coming.
Highlight setting change on the right side of the screen
Evidence supporting changes in the number of transit trips associated with real-time information is limited.
First, it is possible that the “revealed” questions suffered from measurement error and did not capture sufficient levels of detail. For example, the use of trips per week to measure transit travel frequency could be insufficient if a person only makes one or two additional trips per month attributable to RTI. A more reliable way to measure this would be to record trips over an extended period of time (e.g. have respondents report their number of trips per week for all the weeks over the study period). Similarly, a five-point Likert scale for satisfaction may not be sufficiently refined to capture a small increase in satisfaction associated with using RTI.
A second plausible explanation is bias on behalf of the survey respondents when answering the stated questions. The survey methods literature has shown that respondents often have a social desirability bias and will provide an affirmative response that may not align with their actual behavior
Explain bus time, search by route, intersection, etc.
MTA simultaneously launched multiple interfaces and released real-time data openly
The borough-by-borough launch allows us to analyze ridership on routes in Staten Island, the Bronx and Manhattan before and after Bus Time. It also allows for comparison with routes that do not have Bus Time in Queens and Brooklyn.
This roll out by borough – allows us to compare before-after on routes with real-time; also control boroughs (queens & brooklyn)
Likely need more months in Manhattan
Staten island has different demographics