Transforming Transit through Transforming Transit through Insights - - PowerPoint PPT Presentation

Transforming Transit through Insights in Motion Transforming Transit through Insights in Motion

Milind Naphade

Senior Manager, Smarter Cities Solutions Research IBM T. J. Watson Research Center, Yorktown Heights, New York

2

Insights in Motion – Understanding Movement and Optimizing Services Insights in Motion – Understanding Movement and Optimizing Services

Network Data (millions of events/day) Transit System & GIS Data Census & Demographics Data Analytics & Models Smart Fare Card Data (millions of events/day) Information Sources Business Services Outcomes Time of Day Density Maps Origin-Destination Traffic Flow Deep Analysis Planning Large Scale Events, Emergency Response Store Location Siting Transit Planning Location-based Services, Traffic Alerts, Promotions

Reduce Congestion Reduce Journey Time Reduce Carbon footprint Improve Store Traffic of Customers Improve Revenue Reduce Operating Expenses Reduce Emergency Response Time

Insights in Motion Mobility Model Insights in Motion Mobility Model

• Individual and Group Mobility Model
• Location and movement pattern (space, time)
• Meaningful location detection
• Meaningful location classification
• Trip purpose
• Estimated Duration of stay
• Estimated Duration of travel
• Mode of travel
• Calling patterns
• Detecting tourist patterns
• Detecting student patterns
• Estimated demographic profile of user of phone
• Anomalies in regular patterns
• Supply Demand Gap Analysis
• Bus Route Optimization for Small and Medium sized Cities
• Feeder Route Optimization for Multimodal Transit

Impact of Route changes on Jule Transit Impact of Route changes on Jule Transit

Increase in Length of the trip & not designing to action areas

Decrease in Ridership Bigger head ways Less Reliability Increase in

• perating

costs Less Fare Box Less Frequency Less Frequency Negative Perception Negative Perception Few funds to improve system Few funds to improve system Reduction in Federal Funds

4

Process to Improve Jule Transit Process to Improve Jule Transit

Optimize Transit Routes Optimize Stop Placement Contrast Supply vs Demand Optimize Operations Measure unmet demand Suggest new bus routes Time of Day Activity Based New Service area & Demand Census Data Traditional Surveys Online surveys Data gathering using technology X X X Design new routes Redesign services by time of day and activity Create new marketing plan

5

Process to Improve Jule Transit Process to Improve Jule Transit

Insights in Motion

Optimize Transit Routes Optimize Stop Placement Contrast Supply vs Demand Optimize Operations Measure unmet demand Suggest new bus routes Time of Day Activity Based New Service area & Demand Census Data Traditional Surveys Online surveys Data gathering using technology X X Design new routes Redesign services by time of day and activity Create new marketing plan

6

Data Cleaning

Smartphone (GPS) Census Points of Interest

Meaningful Location Detection O/D Estimation Trip Segregation Supply Model

Transit & GIS

Demand Model Clean Sheet Route Optimization Gap Analysis

Telco Network Data

Duration of Stay Estimation Trip Purpose Estimation

Smart Fare Card (RFID)

Trip Mode Estimation Airsage Proprietary Analysis of Telco Network Data

Optimal Routes

Insights in Motion

Process to fix it Process to fix it

Phase 1: Volunteers for Devices Phase 1: Volunteers for Devices Phase 2: Data Collection & Analysis Phase 2: Data Collection & Analysis Phase 3: Route Optimization & Implementation Phase 3: Route Optimization & Implementation Phase 4: System Calibration Phase 4: System Calibration

7

Phase 1 Phase 2 Phase 3

Insights in Motion

Phase I:Devices Phase I:Devices

The goal is to eliminate active user input, and automatically identify travel mode and trip purpose by using mobile devices and information techniques

Smart phones (Androids & Blackberries) are used to provide location, acceleration and route used by time of day. Sample size : 1,000 Volunteers Radio Frequency Identification Device (RFID) are used to capture transit trips. Sample size : 500 Volunteers Cell Tower Data has been acquired from Airsage. Sample size : 15,000+ phones for 3 months

8

Insights in Motion

Phase I:Recruitment Process Phase I:Recruitment Process

9

Smart Phone

Informed Consent Documentation + $10 study participation incentive Recruitment Methods:

• Point of Sale partnership with local cellular agents
• Employer and Campus Events
• General/Community Events

RFID

Free rides on Jule Transit during study period Recruitment Methods:

• Community outreach events, press releases, and email marketing
• On bus outreach to existing transit users

Insights in Motion

Phase I:Data from Smart Phones Phase I:Data from Smart Phones

Acceleration Speed

• Backend

– Setting up cloud based GPS data gathering – Receive Shape file data from city – Receive link for dynamic alerts to be provided to consumers – Hosting of application (for OTA installs)

• Application

– Blackberry platform – Android platform

• Pull-based interaction

– Application anonymously uploads location data – Battery-optimized sampling – Alerts and messages pulled by application from backend

10

Insights in Motion

Phase I:Data from RFIDs Phase I:Data from RFIDs

Acceleration Speed

11

Insights in Motion

Phase I:Data from Cell Tower Data Phase I:Data from Cell Tower Data

Acceleration Speed

• Cell tower data properties
• TAZ zone based
• Include both call, 3G data and

roaming records

• People flows between 7-9 AM using

cell phone call data

• Regions represented by centroids
• Volume represented by line

thickness

12

Insights in Motion

Mode of Transportation and purpose Mode of Transportation and purpose

High Low High Vehiclebus Vehiclebus Or Vehiclecar Median Bicycle Low Walk Zero Static Acceleration Speed Trip Purpose Definition HBWork The trips from home locations to

• ffice locations.

NHBWork The trips from locations other than home to office locations. HBSchool The trips from home locations to school locations. NHBSchool The trips from locations other than home to school locations. HBShop The trips from home locations to shopping areas. NHBShop The trips from locations other than home to shopping areas. HBOther The trips from home locations to

• ther locations.

NHBOther The trips from locations other than home to other locations.

13

Insights in Motion

Phase II: Trip purpose distribution check Phase II: Trip purpose distribution check

Acceleration Speed

Start Time End Time Purpose

7:27 7:39 HBW 8:35 8:47 NHBO 9:30 9:55 NHBW 10:56 11:15 NHBO 11:23 11:31 NHBW 12:22 12:33 NHBO 12:53 13:09 HBW 14:00 14:04 NHBO 14:29 14:48 NHBW 19:28 19:47 NHBO

Points of interest: Businesses, retail, hospitals, schools, public buildings, etc.

14

Insights in Motion

Phase II: Data from Radio Frequency Identification Device (RFID) Phase II: Data from Radio Frequency Identification Device (RFID)

Acceleration Speed

The data is collected from August 2011 – April 2012. There are 43,025 RFID traces with 5,019 RFID traces with duration less than 5 minutes. Moreover, there are 3,002 RFID traces with duration exactly equal to 60 minutes and 35,004 RFID traces with duration >=5 minutes and < 60 minutes; 468 unique RFID tags.

15

Insights in Motion

Phase II: Smart Phone Analytics Validation Phase II: Smart Phone Analytics Validation

• #Volunteers maintaining detailed diaries: 7
• Duration of diaries: 7 or more consecutive days
• Accuracy of detecting meaningful location visited: 93%
• Average distance between detected vs. actual home: 0.06 miles
• Average distance between detected vs. actual work: 0.25 miles*
• Accuracy of trip detection: 96%
• Larger number of trips in diaries occur: In the afternoon

16

* Most of this error is due to the mismatch between “GPS coordinates of the postal address of work versus actual location of entry vs. exit

Insights in Motion

Phase II: Trip Statistics based on Smart Phone O/D Phase II: Trip Statistics based on Smart Phone O/D

17

Sample trip distribution and O-D statistics for Smart Phone data.

Insights in Motion

Trip Statistics based Cell-based O/D Trip Statistics based Cell-based O/D

Sample trip distribution and O-D statistics for cell tower data.

18

Cell-based O/D

Insights in Motion

Phase III Phase III

Route Optimization & Implementation

) constraint routes

• f

numbers maximum ( ) constraint length (trip ) constraint size (fleet ) constraint factor (load * ) constraint y feasibilit (headway s.t. * * * * * min

max max min 1 1 max max max min 3 1 2 1

R M R r D D D R r W h T N R r L P h Q L R r h h h d d d C h T O C C t d t d C z

m r m M m r r M m r m r r r m r N i N j TR tr tr ij N i N j DR r r ij N i N j ij M m r r v v N i N j TR tr tr ij r ij N i N j DR r r ij r ij

m m m m m m m m ij ij m m m m ij m ij m m m

                                          

           

                

19

Insights in Motion

Phase III: Route Optimization with Smart Phone Data Phase III: Route Optimization with Smart Phone Data

Acceleration Speed

Current Routes

Clean sheet optimization to minimize opex, unmet demand and travel time Constraints include fleet size, max transfers, duration, etc.

Clean Sheet Optimal Routes

20

Current Routes

Insights in Motion

Phase III: Route Optimization with Telco Data Phase III: Route Optimization with Telco Data

Acceleration Speed

Current Routes

Clean sheet optimization to minimize opex, unmet demand and travel time Constraints include fleet size, max transfers, duration, etc. Optimal routes can

• reduce OPEX cost up to 40%
• reduce unmet demand by 37%
• reduce avg. travel time from 37 minute average to 10-22 minute average

Clean Sheet Optimal Routes

21

Insights in Motion

Phase III: Pilot Routes Phase III: Pilot Routes

Details of Nightrider

• Focused on unmet evening ridership and

college students.

• The route was designed based on random

survey of students.

• Adjustment to the route will be based on the

Smart phone data provided by student population.

• Further this route will be adjusted based on

final O/D data.

Details of Midtown Loop

• Focused on existing fixed routes
• The route is designed to reduce headways.
• The route is in process of getting

implemented.

• This route will be adjusted in future by

based on final O/D analytics.

22

Insights in Motion

Phase IV: System Calibration Phase IV: System Calibration

Boarding Data Boarding data >= Target ridership

Smart Card rider Smart Card reader Ranger Wireless provider Backend Server

User Data Time of loading Smart Card Loading location Smart Card Reloading location Location Data User ID Usage Accounting Analysis to determine potential ridership Age Income Vehicle

• wnership

Location of Bus stop

TAZ Marketing

NO

Continue Marketing

YES

23

Did the ridership increase after change in marketing

Adjust the route

YES NO

Istanbul in Motion Istanbul in Motion

Objectives

• Create people movement models with anonymous telco data
• Utilize anonymous mobile phone location information
• Build demand side models
• Work with ULASIM AS to evaluate applications of models
• Evaluate existing multimodal transit system use against overall demand
• Explore opportunities to optimize multimodal transit coordination based on gaps
• Deliverables
• Trip Frequency Tables
• Trip Distribution Tables (Origin Destination Matrices)
• Snapshots of zonal occupancy
• Analysis of multimodal transit use against the backdrop of overall movement demand
• Preliminary results on feeder routes for M4 line for all stations
• Outcomes
• First rich large scale movement model level understanding of how Istanbul moves
• Deliverables being used by ULASIM Istanbul to plan feeder bus routes for all stations
• Deliverables will be used by all Istanbul municipal agencies in planning beyond ULASIM

Istanbul.

Results – Population Density and Traffic Snapshot Results – Population Density and Traffic Snapshot

Results – Origin Destination Results – Origin Destination

Trip Analytics: Identifying Meaningful Locations Trip Analytics: Identifying Meaningful Locations

Where People Live Where People Work

Istanbul Movement Analysis w. Vodafone network data

• 4.7 million phones w. 3B+ events/week
• Accurate detection of home, work & meaningful locations

Trip Analytics: Understanding home-to-work commute patterns Trip Analytics: Understanding home-to-work commute patterns

How far people travel from home zones to work How far people travel to come to work zones

Results – Commuter Pain Index Results – Commuter Pain Index