(Center Identification Number: 79063-19)
Sean J. Barbeau, Ph.D.
Principal Mobile Software Architect for R&D
National Center for Transit Research (NCTR) at the
Center for Urban Transportation Research (CUTR)
University of South Florida
4202 E. Fowler Avenue, CUT 100
Tampa, Florida 33620-5375
Phone number: (813) 974-7208
Email address: email@example.com
Multimodal transportation options such as transit, bike, walk, transportation network companies (TNCs
e.g., Uber, Lyft), carshare, and bikeshare are vital to supporting livable communities. To build safe and
effective multimodal infrastructure, Departments of Transportation (DOTs), Metropolitan Planning
Organizations (MPOs), and transit agencies need quality data about how the public is currently traveling
via these modes. However, current data collection techniques for multimodal travel behavior have
limitations which restrict the ability to solve significant real-world multimodal problems.
One example area lacking robust multimodal data is the relationship between (Transportation Network
Companies (TNCs) and public transit. Some see TNCs as a competition to public transit that is primarily
responsible for trends of dropping transit ridership. Others see TNCs as vital first/last mile service that is
a complement to public transportation. Some agencies, such as Pinellas Suncoast Transit Authority
(PSTA) in Tampa, FL, have formed partnerships with a TNC (e.g., Uber) in order to help fill the first/last
mile gap. Other agencies such as Hillsborough Area Regional Transit (HART) have started operating their
own TNC-like service (in the cases of HART, via a contractor Transdev). However, it can be extremely
difficult to capture travel behavior data that includes holistic information about TNCs and public transit
– as of today, there is little hard data that includes origin/destination trip data for transit and TNCs,
especially when one mode is used in place of another. The primary method of capture, on-board
surveys, would completely miss travelers that opted to use a TNC instead of transit for a trip. And, any
information that is captured only covers a day or two of behavior – on-board surveys do not capture
longitudinal behavior. As a result, practitioners and researchers have yet to understand the precise
relationship between the two modes.
Transit rider personal safety and travel patterns also aren’t adequately captured using existing data
collection tools. A 2014 report conducted by the Florida Department of Transportation in collaboration
with Florida’s Transit Safety Network (FTSN) determined that while bus operator assaults are significant
in Florida, assaults on transit vehicle riders were even more significant both in terms of the number of
riders injured and the rate at which these assaults were occurring. The report also stated that more
data is needed to better understand this issue. Current public transportation on-board surveys may
occur face-to-face with riders at a single location during the day, and may not adequately capture how a
rider feels about safety at other locations that they visit at different times of day (e.g., night). As a
result, a transit agency may not have adequate data to respond to concerns about safety on their
system based on quantitative or qualitative information linked to specific bus stops or routes.
Collecting travel behavior data from bicyclists is also challenging. For example, a 2014 USDOT-UTC
Pedestrian/Bicycle Workshop1 determined that a lack of data on when and where bicyclists travel, as
well as their interactions with vehicular traffic, is one of the greatest limitations to better understanding
Florida’s extremely high bicyclist and pedestrian fatality rates. As a result, one of the greatest research
needs is to develop better tools to collect data from multimodal travelers. The Florida Pedestrian &
Bicycle Strategic Safety Plan (PBSSP) emphasizes the need for more reliable and effective data collection
methods (Section 3.5.1).
There have been past efforts to collect bicyclist behavior data via smartphones. San Francisco created
an open-source project “Cycle Tracks” (http://www.sfcta.org/modeling-and-travelforecasting/
cycletracks-iphone-and-android), a mobile app which was used to collect bike path data
from bicyclists’ smartphones. A similar project, Cycle Atlanta (http://cycleatlanta.org/), was
implemented in Atlanta, GA, and was based on the Cycle Tracks open-source code. Strava, a company
that offers a recreational bicycling activity recording app, offers bike path data collected by its users to
FDOT and city planners for a fixed time period for a fee. However, all of these systems only collect bike
path data from bicyclists – they do not collect information for transit or any other modes of
transportation, including connectivity to transit. Additionally, users must start the app just for the
purpose of recording their trip, which is burdensome to the user and can result in fatigue and reduced
data contributions. Also, Strava, the app with the largest number of users, does not provide trip origin-destination
(O-D) data at an individual level.
Other mobile apps have been designed specifically to replace travel behavior surveys, including TRACIT,
Future Mobility Survey, Quantified Traveler, and SmarTrAC, and Florida Trip Tracker. However,
these apps also suffer from user fatigue for manually recorded trips and have only been deployed in
small research settings. With the exception of TRAC-IT, they also do not provide an ongoing incentive or
immediate value to the user for continuing to use the app. Because of user fatigue, these apps are also
typically only deployed for a short time period of several days to a week, resulting in a limited view of
Naturalistic Bike Studies, such as those conducted by Virginia Tech (http://bit.ly/VTTI-NatBikeStudy) and
CUTR (http://bit.ly/CUTR-NatBikeStudy), focus on outfitting bicycles with additional equipment such as
on-board computers and cameras to collect more data about a bicyclist’s first-hand experience.
However, these systems can be very costly to deploy on a large scale, and logistically difficult to manage.
Purpose and Benefit of Research
This research will develop and deploy a system that will collect multimodal travel behavior data on an
ongoing basis directly from users of a popular mobile app for multi-modal information, OneBusAway
(OBA). OBA is currently deployed in over 10 cities around the world, including at Florida transit agencies
Hillsborough Area Regional Transit (HART) and Pinellas Suncoast Transit Authority (PSTA). The purpose
of this research is to increase quality and cost-effectiveness of multimodal travel behavior data
collection. Better data will assist planners in understanding how and where people are traveling via
non-single occupancy vehicle (SOV) modes, which will enable DOTs, MPOs, and transit agencies to
better prioritize infrastructure investments and make operational improvements.
The data collection tool, OneBusAway TRACIT (OBA-TRACIT) plugin, will be developed as an open-source
software library that could be integrated into any mobile app, instantly transforming any existing user
(with their permission) into potential contributors of travel behavior data. As part of this project, OBATRACIT
plugin will be integrated into OneBusAway, instantly adding the ability to collect travel behavior
data from over 325,000 users of OBA nationally. Data shows that the OneBusAway mobile app is very
“sticky” and retains approximately 67% of users that download the app8, compared to an average
retention rate of 20% for other mobile apps9. Multimodal travel behavior data collected from users of
the OneBusAway app (via the OBA-TRACIT plugin) would be immediately accessible by all regional
agencies, resulting in a common view of the data. Additionally, users have an incentive (real-time transit
information) to continue to use the mobile app, and therefore data collection could be ongoing for
months or even years. Additionally, as new regions continue to deploy the open-source OneBusAway
mobile app, travelers in those regions will become potential data contributors.
The versatility of this tool and enormous potential audience of data contributors will enable new
research opportunities and many different types of travel behavior studies well beyond the end of this
project. Travel path data can be passively collected from travelers via various technologies in the mobile
device (e.g., GPS, WiFi, cellular, accelerometers). This data can yield information about trip origins and
destinations, dwell and travel times, and even mode of transportation and purpose information via
machine learning and data mining techniques. To obtain qualitative information, very brief targeted
micro-surveys (1-2 questions) for high-priority issues such as safety could be delivered to travelers when
and where they are likely to have more free time – for example, while they are waiting for the bus at a
bus stop. Connections to social media, with the user’s permission, could be leveraged to easily obtain
demographics information without placing a burden on the user.
This research will assist regions in collecting multimodal transportation data to better understand how
and where people are traveling via public transportation, bike, and pedestrian modes, and will help
inform DOTs, MPOs, and transit agencies in multimodal infrastructure and program investments,
including improvements to bike and pedestrian facilities to access transit and transit rider safety. This
research will contribute towards meeting the goals and objectives related to Data Analysis and
Evaluation portion of the Florida Pedestrian & Bicycle Strategic Safety Plan (PBSSP).
The objectives of this project are to improve multimodal infrastructure planning and transit service
quality by increasing the cost-effectiveness and quality of data from all modes of transportation,
including transit, bike, walk, transportation network companies (TNCs, for example Uber and Lyft),
carshare, and bikeshare. This objective is accomplished via the development of a plugin for the
OneBusAway mobile app to collect travel data including origin/destination, transfers, trip travel path,
and travel time data. The tool can immediately leverage the huge existing user base of OneBusAway
instead of trying to create a new user base specifically for a dedicated travel behavior data collection
1. Reduce data collection cost per completed trip based the cost of existing data collection efforts
used by transit agencies, MPOs, and FDOT.
2. Collect travel behavior data from 500 OBA users that opt-in to detailed tracking using OBATRACIT.
Project Kickoff Teleconference
The principal investigator will schedule a kickoff meeting that shall be held within the first 30 days of
task work order execution. The kickoff meeting will consist of a webinar at least 30 minutes in length.
The purpose of the meeting is to review the tasks, deliverables, deployment plan, timeline, and
expected/anticipated project outcomes and their potential for implementation and benefits. The
principal investigator shall prepare a presentation following the template provided at
The project manager, principal investigator, and research performance coordinator shall attend. Other
parties may be invited, if appropriate.
Certain tasks for this project may occur in parallel, and will be iterative in nature.
Task 1. Design and implement software to collect travel behavior data from OneBusAway users
The research team will identify and prioritize the travel behavior data to be collected from users as part
of this project. Feedback from local transit agency, FDOT district, and MPO will be solicited. The
OneBusAway community and OneBusAway Board of Directors will also be consulted to ensure that
selected data and collection processes aligns with the overall goals of the OneBusAway project,
including user privacy protection. The USF Institutional Review Board procedures for data collection
from human subjects as part of a research project will also be followed, including the filing of proper
documentation with USF as required.
The software that will collect this data from OneBusAway users and submit it to a server for data
storage and processing will be designed and implemented. Various methods of data collection will be
examined based on the type of data being collected (e.g., background software execution for travel path
collection, use of machine learning-based techniques to identify mode transitions).
Task 1 Deliverable:
Software source code to collect data from OneBusAway users uploaded to the opensource
site Github, as well as documentation summarizing the data that is being collected, and any IRB
Task 2. Design and implement software to receive and visualize data from OneBusAway users
This task will design and implement the software that will receive collected travel behavior data from
OneBusAway users, and visualize this data. Feedback from local transit agency, FDOT district, and MPO
will be solicited. This task will include a review of methods and tools to view the data, including the
open-source software projects Traccar (https://www.traccar.org/) and OpenGTS
(http://opengts.sourceforge.net/). When possible, open-source software, existing open specifications
and protocols, and existing sites will be leveraged.
Task 2 Deliverable:
Software source code and/or documentation to receive and visualize data collected
from OneBusAway users uploaded to Github.
Task 3. Deploy and test travel behavior data collection software
Deploy and test the travel behavior data collection software as part of OneBusAway. The travel
behavior data collection plugin will be deployed to a smaller number of users for an initial test (around
200). When it is determined that the travel behavior data collection plugin is working properly based on
feedback from the test users, it will be made available to all OneBusAway Android users.
Task 3 Deliverable:
Information regarding the test deployment, including the number of enrolled users,
duration of enrollment, and amount of data collected.
Task 4. Draft Final and Closeout Teleconference
The PI will submit a draft final report to NCTR administrators.
The draft final report will contain a summary of the software developed as part of this project, and
example data collected during the project.
The draft final and final reports must follow the Guidelines for University Presentation and Publication
of Research available at http://www.fdot.gov/research/docs/T2/University.Guidelines.2016.pdf.
The report must be well-written and edited for technical accuracy, grammar, clarity, organization, and
Thirty (30) days prior to the end date of the task work order, the principal investigator will
schedule a closeout teleconference. The principal investigator shall prepare a Powerpoint presentation
following the template provided at
. At a minimum, the principal investigator, project manager, and research performance coordinator shall
attend. The purpose of the meeting is to review project performance, the deployment plan, and next
Task 5. Final report
Upon approval of the draft final report, the PI will submit the Final Report on two (2) CDs.
Both CDs shall contain the report in PDF and Word formats. CDs must be labeled in a professional
manner and include contract number, task work order number, project title, and date.
The final report is due by the end date of the task work order and should be sent to NCTR
Sean Barbeau, Principal Investigator
Graduate Student TBD, Software Developer & Data Analyst
Phil Winters, Review deliverables
Christine Epps, Project Coordination
It is anticipated that the research project outlined in the scope of work will be completed within 12 months.