DAIYING CHEN DAVID LEVINSONADDAM HALL KAREN MESKOLISA HALL EI EI MIN THUNOLAN LEAVITT SUDHEER SOMESHWARA

Fall 2009

Heinz College, Carnegie Mellon University

a a real-time pilot for the CMU Shuttle

Agenda

History Planning Implementation Results Going Forward

Stakeholders

Starting Problem Advancing knowledge

within CMU community, in line with Traffic21

Benefiting area residents and commuters

Make significant and substantial contributions to public policy and non-profit management

Stakeholders

Our Changing World

Our Solution: Real-Time Information

Real-Time Transportation Information Cutting-edge technology Novel solution to reliability problems

Many benefits To riders To transit providers To community

Agenda

History Planning Implementation Results Going Forward

Deliverables

Port Authority Technical Capabilities Report Public Transit Ridership Surveys myRide website - http://myride.heinz.cmu.edu Funding Request for permanent system Future coursework plans Android Phone GPS tracking application Google Transit Feed Specification compliant database Mobile Webpage Project Document Report

Benchmarking

University of Michigan TransLoc

~60-bus fleet covers 10 routes~Magic Bus was designed by students~Maintained by staff and students~Funded by Transportation Dept.

~Newer company based in Raleigh~Provides services for 15 schools, including Princeton, Auburn, and Yale

Internal CMU Ridership Survey

Goal: identify the most effective and desireddissemination methods for the CMU shuttle

Small, N= 51 Conducted in person at CMU Shuttle stops and on

the Shuttle. Time frame: Weekdays at various times in mid-

October.

CMU Ridership Survey Takeaways

Shuttle riders do have issues with the timeliness of service.

A wide range of people use the shuttle. Shuttle riders have very high levels of access to

Internet and Text plans. iPhones would not be the most effective way to

reach the largest number of people. Focus on a webpage that can be viewed on

mobile devices.

Pittsburgh Community Survey

Goal: Measure attitudes and perceptions in regards to public transit and technology.

Key factors we wanted to measure: Ridership habits Factors affecting demand elasticity for public transit Access to information dissemination methods Receptiveness to various real-time services Perceived value of a real-time system

The questions posed to respondents were modeled after a series of questions used in a 2006 study by the FTA in estimating benefits of a real-time system.

Source: Real-time Bus Arrival Systems Return on Investment Study. Federal Transit Administration, 2006.

Pittsburgh Community Survey Methodology

Our survey was limited in breadth and depth by a limited timeframe and limited resources. The sample size is not intended to be arandom sampling of Allegheny County residents; instead, it attemptsto measure riders and advocates in the Oakland-Downtowncorridor.

N=148 Survey conducted in-person and online 31% Random sample of pedestrians and bus riders in the

Oakland corridor and downtown 35% Students, faculty and professionals in the Higher Education

field 34% Developmental, cultural and transportation advocacy groups

via text message via an automated telephone system

via a standard website via a website for mobile devices/

smartphones

via a smart phone application

30.2

11.1

26.7

8.9

28.3

32.6

22.2

11.1

22.2

13

7

8.9

46.7

33.34.3

16.3

17.8

8.9

22.2

30.4

14

40

6.713.4

23.8

Respondents ranking of dissemination methods1 (Most Prefer) 2 3 4 (5) Least Prefer

Pittsburgh Community SurveyPreferred Delivery Methods

97.1% 90.1% 72.8% 21.3%Access to method:

Pittsburgh Community SurveyPerceived Value

56.5 %34.8 %

6.5 %

2.2 %

Percentage of respondents who felt that "real-time bus information that was accurate, free and accessible

to me would...."

Increase Ridership

Greatly Increase Ridership

No effect on Rid-ership

Negative effect on Ridership

Pittsburgh Community SurveyPerceived Value

Improve my perception of Public Transporta-tion

Make Public Transit more accessible

Make my commute or trip easier

Make Public Transit more appealing

Improve our region's reputation

97.9 %

89.4 %

85.1 %

76.6 %

57.4 %

Percentage of respondents who felt that a real-time Bus System would...

Pittsburgh Community Survey Takeaways

When compared with other metro regions, the Oakland-Downtown corridor has:

The FTA estimated that a system wide real-time system would increase ridership by 6%-8%.

National Average (2006)

Community Survey

Access to Internet 53.3% 72.8%

Access to Mobile Internet 7.2% 21.3%

Dissatisfaction w/ On Time Performance

42.8% 58.4%

Dissatisfaction with Frequency of Service.

51.3% 60.5%

Source: Real-time Bus Arrival Systems Return on Investment Study. Federal Transit Administration, 2006.

Web server

Transmitting Real-time Bus Location Part

Accessing Real-time BUS Location Part

Bus with GPS

Send GPS data to Web serverWeb Application

Mobile Web

G-phone & T-mobile

Riders

Scope Framework

GTFS Data Schema

Estimated Time ModuleLocation Retrieval Module

Map Plug In

Use Case Diagram

Driver

Transport Admin

RiderGeneral User

Start auto-GPS transmission for any Route

Stop auto-GPS transmission for any Route

Change the Route

myRide SystemAdd new Alert for

riders

Add another Admin user

View Current Bus location on

the map

View full schedule for each route

View myRide on

their Mobile Phone

Use Twitter to follow, share the updates

View estimated

arrival time for their bus

stop

Agenda

History Planning Implementation Results Going Forward

Logistics

Graphic Design

Graphic Design

http://myride.heinz.cmu.edu

Marketing Roll-out

Demo: http://myride.heinz.cmu.edu

Highlighted tables are GTFS-compliant schema Improve scalability and future enhancement with Google

GTFS-Compliant Database Schema

Data Source Challenges

Bus stop information not available Collected bus stop information Obtained GPS longitude/latitude from Google Maps

Collaborated with drivers to get accurate schedule Route and schedule data population

3 Routes 23 Stops 78 Trips 1140 records of Stop-times

Route Stops Population Data

Runs as background service on Google Android Phones Transmits GPS data every 5 seconds Easy to use for different routes User-friendly User Interface (UI) for Shuttle Drivers

GPS Transmission

GPS Transmission Challenges

Learning curve of Android Platform GPS providers

Network vs. GPS satellite provider Adjusting GPS transmission interval

1 minute or 50 seconds or 5 seconds Performance vs. Accuracy

Deployment to real phone Versions crisis

GPS background service challenge Reliability of hidden service Does phone screen lock stop our application?

Get GPSTransmit

to Web Server

Background Service

Main Web Interface

Challenges Behind the Scene

Geographic Information System Calculating distance by Vincenty’s formula with

ellipsoidal model of earth Accuracy within 0.5mm[1]

Route distances vs. straight-line distance Mapping raw GPS to nearest bus stop

Geocoding with Google Map Reverse Geocoding Encoded Geopolyline mapping Ajax and timer for updating real-time

Cross-Browser Compatibility[1] Source: http://www.movable-type.co.uk/scripts/latlong-vincenty.html

Challenges: Estimated Time Prediction

Inaccurate schedule stop times

Exponentially Weighted Moving Average Problem with frequent stop times

Kalman-Filter Prediction Algorithms[1]

Consider dwelling times

Various Scenarios Select stop time Schedule time Last trip

[1] Source: Prediction Models of Bus Arrival and Departure Times, University of Toronto

Challenges: Estimated Time Prediction

Is Schedule running?

Get Latest GPS data

Is GPS data

outdated?

Get Next Schedule Time

Get the distance and speed to selected stop

Yes

Yes

No

Display Not Running Now

No

No speed? Or Morewood is in between?

Yes

Predict time

No

Last Trip of the day and passed by?

Display Location without Time

Yes

Mobile Phone Interface

Challenges

Display and bandwidth limitations Layout changes for mobile Decrease page load time

Request redirection Device detection

Request

Users

Transport Admin Interface

User Location Detection

Detecting nearest stop based on user’s current location

Google Gears – Geolocation API

Agenda

History Planning Implementation Results Going Forward

Test Cases

Test Reports

Written Time Stamp

Written Location Desc

Server Time Stamp

Server Location

Notes

2:16 Morewood Garden

2009-11-08 14:18:09

Morewood Ave 1076

2:18 Craig and Forbes Corner (Traffic)

2009-11-08 14:19:28

Forbes Ave 4899

100m

2:19 Forbes and Dithridge

2009-11-08 14:21:23

Forbes and Dithridge

50m

2:20 Webster Hall

2009-11-08 14:12:00

S Dithridge St 155

10m

Test for Route AB – by Ei Ei Min Thu 11/08/09Procedure: attached the phone on bus window without interaction. Phone is charged with laptop on.

Web Counter

Thanksgiving Holiday

Accomplishments

Android deployment GIS (Geographic Information

Systems) Challenges Estimating bus arrival time Mobile Compatibility

Agenda

History Planning Implementation Results Going Forward

Next Steps: Future Enhancements

Enhance System Admin module to update the Route and the stop times on the UI

Improve the Estimated Time algorithm Incorporate the CMU Escort and PTC Shuttle

Route Add advertisements and school announcements

on the website

Next Steps

Continue the Pilot Install more robust hardware Create an iPhone application

Devise traffic monitoring system based on sensor and server capabilities

Pursue funding opportunities Advertisements University Funding Traffic21

Acknowledgements

Robert Hampshire (team advisor) (donation of G1 Phones) CMU Shuttle: Lt. Gary Scheimer, Jim Heverly, Jim

McNeil, Colton Brown, James Collins, & Jason Brown

Ramayya Krishnan, Rick Stafford, Dave Roger, Steve Bland, and Joe Hughes (advisory board)

Hillman Foundation Gary Franko (design and printing support)

Contact Information

Addam Hall (project manager): aehall@andrew.cmu.edu

Ei Ei Min Thu (IT manager): eiei@cmu.edu Robert Hampshire (advisor): hamp@cmu.edu