1. Deducing Mode and Purpose from GPS Data Peter Stopher, Jun Zhang and Eoin Clifford Institute of Transport and Logistics Studies The University of Sydney and Camden FitzGerald Parson Brinkerhoff, Sydney April 2007

2. Deducing Mode and Purpose from GPS Data -- Case Studies 2 Introduction  Increasing use of and interest in GPS devices to measure travel  Used primarily to validate standard diary surveys  Also used in Australia to evaluate VTBC interventions  Recent developments have produced small and lightweight personal devices that are very sensitive

3. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 3 Introduction  GPS collects accurate data on:  Location at each instant  Time at each instant  Speed of movement at each instant  Heading  Data quality measures  Data can be collected as often as every second or as long as desired

4. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 4 Introduction  For transport planning applications:  Logging devices are all that is required  Transmission in real time is not advantageous and may be expensive  4-8 Mb of on-board memory will collect second-by-second data for 1-3 months

5. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 5 Introduction  To substitute for conventional surveys, additional data are needed:  Mode of travel  Purpose of travel  GPS cannot collect these  With certain supplemental data, these can be determined

6. Mode and Purpose Identification

7. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 7 Supplemental Data Requirements  For mode:  GIS of the street network  GIS of all public transport routes (including rail and subway lines)  For purpose:  Locations used frequently by household members  A GIS of the land use at the parcel level

8. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 8 Frequently Visited Locations  Our surveys collect:  Address of each workplace for each household member  Address of each educational establishment attended by a household member  Two most frequently-used grocery stores  Home address is known already  These locations are all then geocoded

9. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 9 Preliminary Steps  Assumes that the data have been subdivided into trips  We define a trip end as occurring whenever the device is stationary for more than 120 seconds  Visual inspection is also used to cross-check and correct some trip ends  About 5 percent of identified trip ends are just traffic stops  About 5 percent of actual trip ends are shorter than 120 seconds and not identified by software

10. Mode Identification

11. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 11 Mode Identification  Proceeds in a hierarchical process:  Identify walk trips first – based on maximum speed  Identify rail, ferry, and other off-network modes next – determined by location of path  Identify bus trips next – based on maximum speed and acceleration and beginning and ending on a bus route

12. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 12 Mode Identification  Identify bicycle trips next – household must have bicycles available  If bicycle is available:  Check maximum speed and acceleration  Check that trip origin is home or is a location to which bicycle has already been used  If all of these are acceptable, then trip is allocated to bicycle

13. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 13 Mode Identification  Remaining trips should be car  Check maximum speed and acceleration  Check that travel remains on the roadway network  If these check out, then trip is car, probably driver  Check to see if origin is home or car was used previously to reach the origin  If not, and car is still identified, classify as car passenger

14. Trip Purpose

15. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 15 Trip Purpose  Examine trip end locations and check against frequently-used locations  Following purposes should be evident:  Home-based work  Home-based school  Some home-based shop  Non-home-based work  Non-home-based school  Non-home-based shop (some)

16. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 16 Trip Purpose Trip List (n) End Points (n+1) Private Level Home Address Work Address Education Address Shopping Address … Public Level Shopping Centre Education Institutes Commercial Places Residence Area … Link Match (Geocoded) To Private Level Destination Link Match (Geocoded) To Public Level Destinations (>1) To Public Level Destination (Best) FAIL Trip Purpose Explanation * search radius: 200 meters

17. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 17 Trip Purpose  30 percent of trips are usually home- based work/school  13 percent of trips are home-based shop  About 5-10 percent of trips are non- home-based work, school or shop  About 70 percent of trips are home- based  This process identifies about 50 percent of trip purposes completely  About 35 percent will have either origin or destination purpose identified

18. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 18 Trip Purpose  Remaining trips are examined with respect to:  Duration of stop  Frequency of visits in GPS period  Nature of land use at the trip ends  These provide further identification of about 35-40 percent of other trip end purposes

19. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 19 Trip Purpose  Problems are:  Multi-use parcels  Shopping centres – people may  Use personal services  Eat a meal  Shop  Visit medical facilities  If the purposes are to be split to HBW, HBSchool, HBOther, and NHB – no problem

20. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 20 Example  The following slides provide an example of the processing steps  GPS data are collected by having respondents carry GPS devices with them for a period of time  Devices are retruned to us and the data are downloaded

21. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 21 Data File from GPS Device  Data are stored as binary in the device  Data are downloaded and converted to.csv file using software  Information is stored with filename which includes deployment information  Gives us information on position, time, heading, speed etc for each data point V,07/03/2006,12:58:49,138.509622,-34.843843,500,78,3,3,8.3 V,07/03/2006,12:58:51,138.509650,-34.843825,500,77,3,3,8.3 V,07/03/2006,12:58:53,138.509686,-34.843809,500,65,2,3,8.3 V,07/03/2006,12:58:55,138.509714,-34.843786,500,66,3,3,8.3 V,07/03/2006,12:58:57,138.509732,-34.843779,500,68,2,3,8.3 A,07/03/2006,12:59:00,138.509741,-34.843580,500,26,8,4,11.6 A,07/03/2006,12:59:02,138.509778,-34.843454,500,34,3,4,11.6 A,07/03/2006,12:59:04,138.509805,-34.843373,500,53,3,4,11.6 A,07/03/2006,12:59:06,138.509824,-34.843316,500,53,3,4,11.5 A,07/03/2006,12:59:08,138.509833,-34.843252,500,28,4,4,11.5 A,07/03/2006,12:59:11,138.509879,-34.843183,500,30,2,4,11.5 A,07/03/2006,12:59:14,138.509943,-34.843149,500,38,2,4,11.4 A,07/03/2006,12:59:16,138.509970,-34.843126,500,56,4,4,11.4 A,07/03/2006,12:59:18,138.510016,-34.843098,499,63,4,4,11.4 A,07/03/2006,12:59:20,138.510007,-34.843115,499,72,3,4,11.4 A,07/03/2006,12:59:22,138.510007,-34.843115,499,77,4,4,11.3 A,07/03/2006,12:59:25,138.509970,-34.843149,498,82,2,4,11.3 A,07/03/2006,12:59:30,138.509961,-34.843172,498,84,2,4,11.3 A,07/03/2006,12:59:32,138.509989,-34.843176,497,86,2,4,11.2 A,07/03/2006,12:59:34,138.509970,-34.843188,497,89,2,4,11.2 A,07/03/2006,12:59:37,138.509961,-34.843199,496,93,2,4,11.2 A,07/03/2006,12:59:39,138.509970,-34.843202,496,75,3,4,11.1 A,07/03/2006,12:59:41,138.509961,-34.843206,495,73,3,4,11.1 A,07/03/2006,12:59:45,138.509925,-34.843222,494,76,2,4,11.1 A,07/03/2006,12:59:54,138.509833,-34.843337,492,95,2,4,11.0 A,07/03/2006,12:59:57,138.509824,-34.843373,491,107,2,4,11.0 A,07/03/2006,12:59:59,138.509833,-34.843390,491,110,2,4,10.9 A,07/03/2006,13:00:01,138.509870,-34.843390,490,109,2,4,10.9

22. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 22 GPS Data Processing Procedure Pre-processingTrip IdentificationMode DetectionPurpose Detection GPS Data Base Map Trip Identification Trip Validation Trip Manual Checking General Mode Detect Public Mode Detect Public Transport Network Household Addresses List Public Places List Position Matching & Purpose Detect Convert Format GPS Record Validation Trip List Trip MapFinal Trip List Deployment Information

23. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 23 Data Prior to Editing

24. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 24 Post Editing

25. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 25 With Mode Added

26. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 26 With Purpose Added

27. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 27 Conclusions  Both mode and purpose can be identified from GPS records  Requires supplemental data such as GIS layers  Requires supplemental questions on:  Bicycle availability  Addresses of frequently-used locations  All other information is available from the GPS record

28. April 2007Deducing Mode and Purpose from GPS Data -- Case Studies 28 Questions ? Please use the Microphone.