A webinar discussing research conducted by the Center for Urban Transportation Research at the University of South Florida that focuses on using mobile apps to improve mobility on various modes of transportation.
"Federated learning: out of reach no matter how close",Oleksandr Lapshyn
2015 Transportation Research Forum Webinar - Enabling Better Mobility Through Innovations for Mobile Devices
1. Center for Urban Transportation Research | University of South Florida
Enabling Better Mobility Through
Innovations For Mobile Devices
Sean J. Barbeau, Ph.D.
Transportation Research Forum Webinar – Sept. 30th, 2015
National Center for Transit Research
2. 2
Opportunities
• 97% of US households have mobile phones (Nov. 2014)
[1]
– Avg. household owns 5.2 connected devices
• Smartphones accounted for 66% of total phone
market in 2014[2]
– Estimated to increase to 9 out of 10 phones in 2018[2]
• Mobile app use up 76% in 2014[2]
• National Center for Transit Research (NCTR)
– UTC focused on public transportation, other
alternatives to single occupancy vehicle travel
[1] http://www.ctia.org/
[2] http://www.gartner.com/newsroom/id/2944819
3. 3
Overview
Problems
• Increase riders
independence via fixed
route transit
• Assess impact of real-time
bus arrival information
• Collecting multimodal data
to support community-
based social marketing
NCTR Project
TRAC-IT
OneBusAway
Travel Assistance Device
(TAD)
Funded by National Center for Transit Research
(NCTR) and Florida Department of Transportation
4. 4
Problem – Exiting the bus
• One of the hardest of the 23 skills necessary
for riding public transportation
– Especially for those with intellectual impairment
– Hardest skill to “travel train”
• Paratransit is expensive for transit agencies
– ~10x the cost of one-way fixed route[1]
• Paratransit limits independence and
spontaneity of travel for some riders
[1] National Transit Database (2005)
National Center for Transit Research
5. 5
TADTAD
Cancel Select
Select Trip
(1) Home to Work
(2) Work to Home
(3) Home to Movie
Work to HomeWork to Home
Back #
Distance to Final Stop:
5.6 miles
18 Livingston West
TADTAD
OK
Pull the Cord Now!
(+Sound and Vibration)
Travel Assistance Device (TAD)
• TAD mobile app tells the traveler to “Get Ready” and
“Pull the Cord Now!” when it is time to exit the bus.
• Prompts are visual, auditory, and tactile
Funded by NCTR, FDOT, TRB IDEA program
6. 6
Evaluation & Results
• Partnered with USF Florida Mental Health Institute &
Hillsborough Area Regional Transit
– 3 individuals with moderate intellectual impairments
– Evaluated requesting stop, getting off at correct stop
– ABAB experimental design
– Safeguards in place
• When no prompts were given, all subjects failed to pull
the stop request cord and exit the bus at the correct
location
• When TAD was used, the subjects pulled the stop request
cord and exited the bus at the correct location
• One rider without TAD asked the bus driver for help
– Driver gave the rider incorrect directions
– Rider exited the bus at the wrong stop
Bolechala, Miltenberger, Barbeau, and Gordon. “Evaluating the Effectiveness of the Travel Assistance Device on the Bus Riding Behavior of Individuals
with Disabilities,” 37th Association for Behavior Analysis International (ABAI) Annual Convention, Denver, CO, May 27-31, 2011. Paper #11396.
7. 7
What is OneBusAway?
• What? Suite of tools that provides real-time
bus/train tracking information
– Originally deployed in greater Seattle, WA
– Started as graduate student project at UW, grew
to over 100,000 unique weekly users
• Why? Make riding public transit easier by
providing good info in usable formats
– Research evaluates the impacts
• Includes apps on many platforms:
http://onebusaway.org
+ Web, SMS, IVR,
and more…
8. 8
Problem – Lack of transit apps
• In some cities, transit apps aren’t as prevalent
• OneBusAway is open-source software
• How can we replicate success of OneBusAway in
other cities?
9. 9
Multi-region architecture
• Region
information is
stored in
centralized server
directory
• Apps now find
nearby regions
using Regions API
• Supports multiple
cities!
Barbeau, Borning, Watkins. “OneBusAway Multi-region: Rapidly expanding mobile transit apps to new cities,” Journal of Public
Transportation – Vol. 17 No. 4 (2014).
10. 10
OneBusAway Tampa
• USF lead a pilot deployment of OneBusAway multi-region in Tampa
• Survey lead by Dr. Candace Brakewood and Dr. Kari Watkins at Georgia Tech*
Objective: Quantify the impacts of real-time bus information on transit rider
behavior and satisfaction in pilot deployment prior to public launch
Methodology: Before and after web-based survey with a non-user (control)
group
BEFORE SURVEY
of Control Group
(approx. 230 participants)
AFTER SURVEY
of Control Group
(107 Non-Users)
BEFORE SURVEY
of OneBusAway Group
(approx. 230 participants)
AFTER SURVEY
of OneBusAway Group
(110 OneBusAway Users)
* Dr. Candace Brakewood is now at City College of New York
No
OneBusAway
11. 11
OneBusAway Tampa - Results
• Significant improvements in the waiting
experience[1]:
– Decreases in self-reported usual wait times by ~2
min.
– Decreases in negative feelings, particularly
frustration
– Increases in satisfaction with wait times
• Other research is ongoing
– Affects ridership? (in NYC, yes![2])
– Issue reporting / rider feedback
[1] Brakewood, Barbeau, Watkins. “An experiment evaluating the impacts of real-time transit information on bus riders in Tampa, Florida”,
Transportation Research Part A: Policy and Practice, Volume 69, November 2014, Pages 409-422
[2] Brakewood, C., Macfarlane, G., and Watkins, K. ”The Impact of Real-Time Information on Bus Ridership in New York City.” Transportation
Research Part C: Emerging Technologies, 2015.
12. 12
TRAC-IT
• Allows “high-definition”
view of travel
• Frequent sampling
allows us to determine:
– Path, distance traveled
– Origin-Destination pairs
– Average speeds
• Two modes:
– “Passive” - GPS only
– “Active” - GPS + input
12
13. 13
Battery Life Problems!
• Infrequent tracking
solves energy, data
problems
• BUT, doesn’t give us
the data we want:
– Path, distance
traveled
– Origin-Destination
pairs
– Avg. speeds
13National Center for Transit Research
14. 14
What is “Stationary”?
Detecting User Movement
Moving Stopped d
4 second
GPS
sampling
5 min.
GPS
sampling
• What if we only sample GPS while the
user is moving?
• GPS noise causes uncertainty in states
• Many false transitions waste battery
energy
15. 15
4 second
GPS sampling
5 min.
GPS sampling
State
0
State
1
State
n – 1
State
n
Move directly to state[0] when speed
exceeds high_speed threshold
Location
Recalculation
Interval = 4 sec.
Location
Recalculation
Interval = 8 sec.
Location
Recalculation
Interval = 64 sec.
Location
Recalculation
Interval = 128 sec.
Move gradually towards state[n] when (speed < low_speed value)
and (distance_between_fixes < distance_threshold).
Move gradually towards state[0]
when (speed < high_speed value)
and (distance_between_fixes >
distance_threshold).
GPS Auto-Sleep - Dynamically change the GPS sampling rate
Moving Stopped
U.S. Patent # 8,036,679 – Optimizing performance of location-aware applications using state machines
17. 17
Critical Point Algorithm
• Purpose – to reduce battery energy expenditures and amount of data transferred by
eliminating non-essential GPS data
• Pre-filters real-time GPS data on mobile device before it is wirelessly transmitted
U.S. Patent # 8,249,807 – Method for Determining Critical Points in Location Data Generated by Location-Based Applications
18. 18
Critical Point Algorithm
• Avg. point reduction of 77% per trip
• Avg. 18.8kB saved per trip
• Average distance error percentage under 10%
• On avg., as Tx interval doubles, battery life doubles
Min Max Avg.
5th
percentile
25th
percentile
50th
percentile
68th
percentile
95th
percentile
Total Critical Point Count 2 322 35 3 13 27 38 97
Total GPS Fix Count 20 3,710 193 31 74 130 188 511
% Savings 20.83% 99.40% 77.43% 47.97% 69.49% 80.00% 86.83% 95.84%
Bytes Saved* 595 403,172 18,883 2,380 6,426 12,138 17,493 54,788
Distance Critical Points (m) 0.00 1,043,805.50 7,437.09 328.14 1,162.37 2,675.00 4,049.37 22,815.61
Total Distance (m) 2.36 1,087,043.20 7,878.02 380.79 1,252.55 2,913.39 4,345.91 24,231.34
Distance Error Percentage 0.00% 100.00% 8.90% 1.94% 3.98% 6.20% 8.70% 24.11%
* Based on 119 bytes per UDP payload
U.S. Patent # 8,249,807 – Method for Determining Critical Points in Location Data Generated by Location-Based Applications
19. 19
Measuring Spatial Patterns of
Activity-Travel for Carsharing
Minor Axis
Major Axis
Standard deviation ellipse (SDE)
Y
X
• USDOT Value Pricing Project
• TRAC-IT w/ 30 participants in
Tampa, avg. 40 days per participant
• Daily emails so users can view data
• Results:
– Carsharing users have smaller
activity space (0.5 sq mi) than non-
carsharing (7.8 sq mi), using SDE
– Activity space of carsharing users
contracts while using carsharing (.2
sq mi vs. 0.5 sq mi), but is directed
Concas, Barbeau, Winters, Georggi, Bond. “Using Mobile Apps to Measure Spatial Travel Behavior Changes of Carsharing Users,”Proceedings of 2013
Transportation Research Board Conference, Washington, D.C., January 13-17, 2013.
20. 20
Conclusions
• Mobile devices offer many opportunities, many
challenges
• Multi-disciplinary collaboration yields innovation
• Using technology to measure and impact travel
behavior works!
• Funding sources:
– Florida Department of Transportation
– National Center for Transit Research
– Transportation Research Board (TRB) IDEA program
– FHWA Value Pricing Pilot Program
21. 21
Thanks!
Sean J. Barbeau, Ph.D.
barbeau@cutr.usf.edu
813.974.7208
Principal Mobile Software Architect for R&D
Center for Urban Transportation Research
University of South Florida
http://www.linkedin.com/in/seanbarbeau
Protected under U.S. patents #8138907, 8169342
Funded by the National Center for Transit Research and
Florida Department of Transportation
Editor's Notes
NCTR is a Tier 1 University Transportation Research Center at USF
Age 5 to 15 years
6.3
Age 16 to 64 years
12.3
Age 65 years and over
41.0
Survey monkey
Participants randomly assigned to groups
Incentive: free one day bus pass
Response rate for second survey ~60%
To understand how carsharing pricing might affect travel behavior, this study compares the spatial dispersion of out-of-home activities undertaken by participants carrying GPS-enabled mobile phones with TRAC-IT pre-installed, differentiating between those who used the carsharing program and those who did not use the carsharing program in this period.
To measure the spatial extent of out-of-home activities across the urban landscape, researchers employed area-based geometric measures developed in the field of transportation geography. Different metrics that describe the spatial extent of activity locations can be used. The simplest measure is represented by the standard distance circle (SDC) (or standard distance deviation), which is essentially a bivariate (i.e., showing the relationship between two variables) extension of the standard deviation of a univariate distribution. The coordinates represent longitude and latitude measurement of each activity and are reported in meters following the Universal Transverse Mercator (UTM) coordinate system. To analyze the GPS data from the TRAC-IT mobile application, the GPS latitude and longitude measurements were converted from the World Geodetic System (WGS) geographic coordinate system to the UTM coordinate system using the appropriate transformations.
Activity locations are those visited by surveyed individuals during a specified time interval, in this case the weeks of March 7 through April 22, 2011. Thus, the standard distance of an individual’s activity pattern is estimated as the standard deviation (in miles) of each activity location from the mean center of the complete daily activity pattern. While individuals identified as carsharing users engage in shorter trips (2.6 miles) than all other users (4.2 miles), they relied on carsharing to make longer trips (8 miles vs. 1.7 miles for non-carsharing trips).