Ed Hard Byron Chigoy Praprut Songchitruksa, Ph.D., P.E. Steve - PowerPoint PPT Presentation

New Methods and Technologies for Collecting Origin-Destination (O-D) Data Ed Hard Byron Chigoy Praprut Songchitruksa, Ph.D., P.E. Steve Farnsworth Darrell Borchardt, P.E. 15 th National Tools of the Trade Conference Charleston, South Carolina September 13, 2016

Presentation Overview • Overview of cell, GPS, and Bluetooth O-D data • Takeaways and lessons-learned from prior studies • Capabilities and limitations by technology • Comparisons of O-D data between technologies • O-D data suitability by study type 2

New Technology in Travel Surveys GPS Cellular Data Validation Tools - classification Bluetooth Trip Tables counts and travel Trip Length Freq. time Corridor Studies Select Link and Zone Analyses Residency Status 3

Cell Data Overview 4

Cell O-D Data What it Represents • Estimated device movements/flows • Spatio-temporal analysis of device over weeks/months • Algorithms used to ‒ ID activity points and trip ends based on dwell times, speed, proximity ‒ Impute/derive estimated device trips based on patterns between home, work, other activity locations • Each sampled device computed individually, then expanded, then aggregated with others 5

Cell O-D Data What it Represents • Data from 1 cell provider and 1 carrier (currently) • Possibly a combination of linked and unlinked trips • Analysis of activity points, clusters 6

Cell Data Examples Select Zone Analysis External Surveys • Mobile Bay Bridge Source: Airsagepopulationmovements_v9 7

Cell O-D Data Example Regional Corridor Study Source: Fussel R, Gresham, C. No Horsing Around, A Hole in One with Mobile Phone Data, TRB TPAC Conference, 2015. Through Trips Between Counties Surrounding Moore County 8

Takeaways from Cell O-D Studies Beginning in about 2010, at least 30+ • Majority of studies for trip matrices, modelling • Cell data acceptable, lower cost option for O-D • Total cell derived trips compare well to total model trips, differences surface when disaggregated • Accuracy increases as size of analysis geography increases • Important attributes; larges sample size, basic trip purposes, residents, visitors, commuters • Inability to discern vehicle type or mode 9

Takeaways from Cell O-D Studies Beginning in about 2010, at least 30+ • Not well suited for small urban TAZs due to poor accuracy • Uncertainty on how much to aggregate internal TAZs • Challenges in configuring external cell capture areas • Misses short trips (proportion unknown) • Mixed results on trip purpose (research needed) • Are commercial vehicles under-represented? 10

GPS O-D Data Overview 11

Types of GPS Data • Primary GPS Data – You collect and own raw GPS data – Resource intensive – Limited sample size – Customized algorithm R&D possible  Third Party GPS Data (focus of this presentation) – Obtained from private data providers – Raw data are protected due to privacy – Larger sample size – More economical – Work with data providers for specific customizations 12

Third Party GPS Data Providers, Sources Provider s • INRIX – ‘Insights Trips’ O -D product • HERE – ‘Trip Data’ O -D product • TomTom – No O-D product, but case-by-case • Data continuously collected, compiled; processed for O-D and anonymity Data Sources • In-vehicle navigation systems, fleet/freight, mobile devices, mobile apps, portable navigation devices 13

Caveats for Third-Party GPS Data • GPS traces are available only when users turn on navigation session . • Non-commercial GPS users may not immediately activate navigation session • GPS data sources can be biased towards specific user groups. • Data providers generally apply anonymization techniques in time and/or space to ensure privacy 14

Data Products from GPS Data Providers • O-D table • Trip records • Trip records with waypoints 15

Takeaways from GPS O-D Studies Recent Studies with New O-D Products • New O-D products in 2015/2016 a major advancement • Ability to provide routing between O-Ds using waypoints and/or Traffic Message Channels (TMCs) • Enhanced ability to assess route choices and patterns • Ability to analyze travel patterns of different GPS sources/vehicle categories • Low sample penetration, but improving • Commercial bias, but data can be segregated by sources so can segregate • Driver habits of using (consumer) GPS major influence on data 16

GPS O-D Data Example – Select Link Analysis Source: Paul Morris, SRF Consulting Group, I-94/TH 62 Study, 1/20/16 presentation to Met Council 17

Bluetooth O-D Data Overview 18

Bluetooth O-D Data What is it? • Wireless technology for exchanging data over short distances • BT embedded in mobile phones, GPS, in- vehicle navigation systems • Each BT device has a unique Media Access Control (MAC) ID • Permanent or portable sensors along roads detect MAC IDs • MAC ID’s matched between sensor locations 19

Bluetooth Data What it represents • Samples of ‘portions’ of trips detected by roadside sensors • Does not detect trip ends • Expanded to counts when data collected • Unlike cell and GPS, does not detect trip ends • Not feasible for E-I/I-E trips • Caveat – ability to change MAC address, reducing detection rates 20

Takeaways from Bluetooth O-D Studies • Vetted source for travel movement data for small scale studies (does not provide true O’s or D’s) • Detection rates by study area about 5-10 percent • Detection rates by roadway about 3-12 percent Avg. Daily Avg. Daily Percent Area Sites Count Reads Match Tyler, TX (2014) 20 175,343 12,975 7.4% Austin, TX (2013) 14* 2,219,908 118,365 5.3% Omaha, NE (2013) 22 235,526 10,470 4.4% Corpus Christi, TX (2013) 9 226,098 20,059 8.9% Bryan/College Statiton, TX 13 92,076 7,003 7.6% (2011) 21

Technology Comparisons 22

Comparison of Characteristics by Technology Technology O-D Data Element Cell GPS Bluetooth Cell sighting Data unit based on event or GPS ping MAC address network hand-over Positional accuracy Ave. 300 meters 1-10 meters About a 100 meter range Data saturation/ Varies from about 3 percent to Good, but varies Relatively low 10 percent penetration Varies widely, in Sample frequency In seconds or minutes In seconds minutes to hours Continuous data Sometimes, but typically Yes, but only in about 100 No, random events pieces of trips captures meters range of reader stream? MAC address matches between How trips estimated Based on activity Trip based on GPS data readers. Trip ends cannot be clusters stream and defined determined. IDs scrambled and Encrypted to Removal of digits from MAC time/distance offsets Anonymization anonymize individual address, aggregating data prior applied. Actual trip ends device IDs to O-D table creation not provided 23

Comparison of Cell and GPS Data Points • Small green dots = GPS points • Pink circles = cell device on the move • 5 mile stretch of SH 295 in DC • 16 cell data points Source: airsagepopulationmovements_v9 powerpoint 24

Comparison of Characteristics by Technology Suitability by Cell Third-Party Comments Geographic Scale Data GPS Data GPS more comparable to Bluetooth. Limited ability to  E-E trips limited apply E-E travel time constraints with cell. For large regions cell may be best at this time due to good   E-I/I-E trips sample penetration. However, GPS may be better at urban TAZ and smaller geographies. External Surveys Cell estimates basic trip purposes based primarily on   Trip purpose device’s home and work locations . Research needed here.   Residency Resident vs. non-resident splits needed for some models Commercial/ GPS can split O-D data into freight and non-freight sources   Freight (cars, apps, and freight categories). Route GPS can determine route between O-Ds using waypoints   information and TMCs. Ability to apply   travel time Typically needed to develop E-E trips/matrices. constraint 25

Comparison of E-E Trips By Technology • 2014 Tyler External Study Comparing Cell, GPS, and Bluetooth O-D • Bluetooth used to estimate E-E movements • Bluetooth E- E’s used as benchmark against cell and GPS E - E’s 26

Comparison: Cell and GPS Data for Corridor Studies Suitability by Geographic Third-Party Cell Data Comments Scale GPS Data Within Urban Areas Cell not well suited doe to data lack of   (operational) positional accuracy Within urban areas GPS has better accuracy and can provide Corridor Studies  (planning, select link limited routing. GPS has ability to constrain data to analysis) corridors in urban settings   County to county Cell sample size makes it best for total traffic, plus it can inform on residents, Multi-county metro   visitors, commuters, etc. However, GPS regions needed for freight. Depends on study objectives   Statewide 27