Processing GPS Data from Processing GPS Data from Travel SurveysTravel Surveys

Peter Stopher and Camden Peter Stopher and Camden FitzGeraldFitzGerald

Institute of Transport and Logistics Institute of Transport and Logistics StudiesStudies

The University of SydneyThe University of SydneyQingjian JiangQingjian Jiang

Parsons Brinkerhoff Australia Pty Parsons Brinkerhoff Australia Pty Ltd.Ltd.

IntroductionIntroductionConsiderable recent growth in

applications of GPS in transportCapability of producing huge data

streamsNeed for custom software to interpret

and use the dataDescription of 4 data processing

applicationsDescription of 2 data analysis steps

ContextContextUse of a GIS base for display and

analysisOur selections is TransCAD®

Use of developer tools in TransCAD®: GISDK ®

Equipment used:GeoStats GeoLogger (in-vehicle and

wearable)Neve StepLogger (wearable)

Application 1: Application 1: Converting to Trip Converting to Trip

RecordsRecordsData must first be reformatted

to match input specifications of TransCAD®

Data are usually in one of the standard NMEA streams

GeoLogger and StepLogger provide slightly different data streams

GeoLogger Data StreamGeoLogger Data Stream

VALID LATITUDE

HEMI-SPHERE LONGITUDE

HEMI-SPHERE

UTC TIME

UTC DATE

SPEED (KPH)

HEAD-ING

ALTI-TUDE HDOP

SATEL-LITES

V 33.89243 S 151.19162 E 50655 190202 0 0 43 2.4 3A 33.89237 S 151.19165 E 50658 190202 20 23 43 2.4 3V 33.89205 S 151.19153 E 50703 190202 0 0 43 2.4 0A 33.8916 S 151.19128 E 50708 190202 40.2 340 43 2.4 3V 33.89112 S 151.19105 E 50713 190202 0 0 43 2.4 0A 33.89077 S 151.19088 E 50718 190202 28 324 43 2.4 3A 33.8904 S 151.1906 E 50723 190202 35.4 327 43 2.4 3A 33.88993 S 151.19023 E 50728 190202 45.9 326 43 2.4 3A 33.88963 S 151.19005 E 50733 190202 19.3 351 43 2.4 3A 33.8893 S 151.19028 E 50738 190202 39.9 41 44 2.4 4A 33.88895 S 151.1908 E 50743 190202 45.9 53 42 2.4 4A 33.88863 S 151.1914 E 50748 190202 45.2 58 42 2.4 4A 33.88832 S 151.19198 E 50753 190202 45.5 57 43 2.4 4A 33.88797 S 151.19262 E 50758 190202 55.3 51 43 2.4 4A 33.88747 S 151.19325 E 50803 190202 57.9 43 42 2.2 4

StepLogger Data StreamStepLogger Data StreamValid Date Loc Time Longitude Latitude Altitude Heading Speed Satellites HDOPA 23/04/2005 0:56:50 138.6184 -34.80916 -3 145 1 5 2.3A 23/04/2005 0:56:56 138.6184 -34.80916 1 8 1 6 2A 23/04/2005 0:57:07 138.6184 -34.80909 5 80 1 6 1.5A 23/04/2005 0:57:08 138.6184 -34.80908 5 316 1 5 2.2A 23/04/2005 0:57:12 138.6184 -34.80908 6 118 1 6 2A 23/04/2005 0:57:13 138.6184 -34.80908 6 106 2 6 2A 23/04/2005 0:57:14 138.6184 -34.80908 7 90 4 6 2A 23/04/2005 0:57:15 138.6185 -34.80908 7 90 4 6 2A 23/04/2005 0:57:16 138.6185 -34.80908 8 92 4 6 1.6A 23/04/2005 0:57:17 138.6185 -34.80908 8 95 4 6 2A 23/04/2005 0:57:18 138.6185 -34.80909 9 92 5 5 1.6A 23/04/2005 0:57:19 138.6185 -34.80909 9 86 5 8 1A 23/04/2005 0:57:20 138.6185 -34.80908 9 87 5 5 1.9A 23/04/2005 0:57:21 138.6186 -34.80908 8 99 6 6 1.4A 23/04/2005 0:57:22 138.6186 -34.80908 8 103 6 8 1A 23/04/2005 0:57:23 138.6186 -34.80909 8 92 5 5 2.2A 23/04/2005 0:57:24 138.6186 -34.8091 7 93 4 6 2A 23/04/2005 0:57:25 138.6186 -34.8091 7 98 5 5 1.9A 23/04/2005 0:57:26 138.6187 -34.80911 7 102 5 8 1A 23/04/2005 0:57:27 138.6187 -34.80911 8 100 6 8 1A 23/04/2005 0:57:28 138.6187 -34.80912 8 92 5 6 2A 23/04/2005 0:57:29 138.6187 -34.80912 8 97 5 6 2A 23/04/2005 0:57:30 138.6187 -34.80912 8 94 5 7 1.3A 23/04/2005 0:57:31 138.6188 -34.80913 8 93 5 5 3A 23/04/2005 0:57:32 138.6188 -34.80913 8 89 5 5 1.6

Differences and Differences and SimilaritiesSimilarities

Both record latitude and longitudeGeoLogger includes hemisphereStepLogger uses plus and minus

Both record date and timeGeoLogger is UTCStepLogger converts to local

Both record altitude (not used), speed, and heading

Similarities and Similarities and DifferencesDifferences

Both record HDOP and number of satellites in viewNumber of satellites – 3 is minimum

for 2D solution and 4 for 3DHDOP is horizontal dilution of

precision – larger values indicate poor quality of data (little dispersion of satellites)

Processing DataProcessing DataCalculate local time and local dateCorrect latitude and longitude and

remove hemisphere designations (when necessary)

Calculate:Distance from previous record, using

latitude and longitudeTime in seconds based on zero seconds at

midnightElapsed time from the previous recordActual date in date format and day of

week

Remove Invalid PointsRemove Invalid PointsAll data points with too few (less than

three) satellites in view and/or a value of HDOP of five or greaterExcept for the first occurrence of such

data points in any group of data points of this type

Any data points where no movement is recorded based on: Speed of zeroLess than a 15 metre (0.00005º) change

in either latitude or longitudeHeading being zero or unchanged

Trip DetectionTrip Detection Detect a trip end when:

Ignition has been turned off (in-vehicle device only) Ignition has not been turned off (in-vehicle device

only) Direction is reversed, but there is a very short or

non-detectable stop Detect a trip end from wearable data, where

the device is always on Determine whether a trip ended and a new

one started during a period of signal loss Determine whether there is a loss of signal at

the beginning of a trip and repair the data record

Trip Detection Trip Detection HeuristicsHeuristics

The difference in successive latitude and longitude values is less than 0.000051 degrees AND

The heading is unchanged or is zero AND

Speed is zero ANDElapsed time during which these

conditions hold is equal to or greater than 120 seconds

If ignition is turned off for in-vehicle for 30-120 seconds – probable trip end

Trip Ends for Trip Ends for Wearable DevicesWearable Devices

The GeoLogger works as for the in-vehicle device

The StepLogger requires different logic, because zero speeds are not recordedDefines a Stop Point Set when there is a

group of records within 30 metres of each other

When the Stop Point Set lasts at least 120 seconds, it becomes a potential trip end

When a new point falls outside this set, a new trip is assumed to have started

Completion of Trip Completion of Trip RecordsRecords

Once trip ends have been detectedAll other manipulations of the data are

completedElapsed time and distance for the trip

are obtainedEach trip is given a trip numberEach point in the trip is given a new

sequence number

Application 2: Application 2: Correcting “Cold Start” Correcting “Cold Start”

ProblemsProblemsThese occur when there is a delay in

position acquisition at the start of a trip

For both wearable and in-vehicle devices, this may result in a missed trip, when the trip is short

If acquisition occurs within 100 metres, this is not considered to be a problem

Identification of the Identification of the ProblemProblem

Determine the distance between the start of this trip and the end of the previous one If less than 100 metres, no action is

requiredDetermine in which zone the trip

end liesZone I is the Sydney CBD and

surroundingsZone II is the rest of the metro area

Correction ProcessCorrection ProcessCompute the average speed of the

last 10 records and the first 10 records of the new trip

Using these speeds and the look-up tables for gap speed, we determine a speed for the gap

Compute the distance for the gapUses straight line distance for under

150 metresUses minimum time path distance with

end corrections for longer distances

Time EstimationTime EstimationUsing the speed and distance from

the previous steps, determine the time taken

Trip time also adds in an average amount of terminal time for either Zone I or Zone II

This process works for car tripsA procedure for other trips is still

being developed

Application 3: Signal Application 3: Signal LossLoss

Signal loss occurs in urban canyons, under heavy tree canopies, in tunnels, and in some public transport vehicles

Two issues:Has a stop occurred while the signal

was lost? What path was traversed while the

signal was lost?

Non-Car DataNon-Car Data If the gap occurs in data where the

GPS is NOT in a carDetermine the location of the point of

signal loss and the point of signal resumption

Determine if there is a rail, bus, or ferry path that connects the two

Determine if the connecting path is reasonable for the time between signal loss and signal resumptionIf it is reasonable, then interpolate the route

Data from an Urban Data from an Urban CanyonCanyon

Sydney CBD – Urban Canyon

Car DataCar Data Use the preceding 10 records and following

10 records to compute average speeds Connect the last point before and first

after with a minimum time path Estimate the implied speed along the

minimum time path, given distance and elapsed time

Compare this to the average speed for the last 10 and first 10 records either side of the gap If implied speed is less than 50 percent of

average speed, then assume a stop has occurred

Application 4: Application 4: Displaying the DataDisplaying the Data

After completing processing, trips from one day are stored in one file

For each day, each trip within the day receives a distinct colour code for the day

Trips are converted to solid lines, which are drawn on maps offset, so that each trip on a given link can be seen separately

A table is also created giving primary information about the trip

Raw Data DisplayRaw Data Display

Processed Data DisplayProcessed Data Display

Determining ModeDetermining ModeBegins with defining walk trips

Average speed maximum of 6 km/h and maximum speed of 10 km/h

Also, may not follow street system, especially where parks and open spaces occur

Next, define public transport tripsUsually follow a walk tripMay also be followed by a walk tripAverage speed in the range of 10-40

km/h

Determining ModeDetermining Mode If trip either disappears, or follows

entirely a public transport route, it is defined as public transport Identification of rail is usually made

from station boarding and disappearance of GPS signal

Identification of bus is usually made from speed and route

Remaining trips are bicycle or carDetermined principally by average

speeds

PurposePurposeUses address data collected at

recruitmentWorkplacesSchoolsGrocery stores

Requires a GIS of land use by parcelTrip purposes are identified from

known locations for work, education, and shopping

Other purposes are defined by the land uses

Processing and Data Processing and Data ManagementManagement

All the preceding steps are accomplished in about 2 hours for one week of GPS data

A folder is defined for each device for each householdAll files created for that device are

stored in that folderFiles can be examined at various

intermediate points for review

ConclusionsConclusionsProcessing steps identify about

95 percent of trip ends correctlyData gap repair (cold starts and

tunnels/canyons, etc. are found to be very successful

Mode and purpose identification is about 95 percent

Fully automated processing is still not completely possible

Future DirectionsFuture DirectionsCurrent processing has reduced

the time required for processing to about 25 percent of initial time requirements

Future potentials for better identification of cold starts/canyons, mode, and purpose from fuzzy logic