GPS for Route Data Collection Lisa Aultman-Hall Dept. of Civil & Environmental Engineering University of Connecticut
Acknowledgements Reema Kundu and Eric Jackson University of Kentucky Wael ElDessouki and Jianhe Du University of Connecticut with helpful advice from Jean Wolf, Sean Doherty and Martin Lee Gosselin Aultman-Hall - August 2001
Research Interest Route Choice Behavior NOT how people should route (generating optimal routing) what motivations (particularly beyond travel time) affect routing scenery, traffic control, road type, congestion, turns population segmentation Aultman-Hall - August 2001
Who Cares? measure benefits of ITS such as route guidance and traveler info systems dissaggregate exposure for crash and safety analysis improve traffic assignment models Aultman-Hall - August 2001
Underlying Assumption all drivers seek to minimize their own travel time stochastic algorithms account for variation Aultman-Hall - August 2001
Aultman-Hall - August 2001 Travel Route Data
Wide-spread route data is now within reach computer power and memory Geographic Information Systems (GIS) Global Positioning Systems (GPS) Aultman-Hall - August 2001
Satellites and receivers are synchronized so they generate the same code at the same time. Receivers know the satellite orbits. Product of time difference and speed of light is distance.
But We have more than one satellite overhead
Typical Errors satellite clock ephemeris receiver atmosphere multipath S/A 2 feet 2 feet 4 feet 12 feet? up to 25 feet Multiply these values by PDOP to get the real time position accuracy. Good PDOP s range from 4 to 6. Aultman-Hall - August 2001
Vehicle + GPS Receiver fleet location and management in-vehicle navigation route data!!! Aultman-Hall - August 2001
A Model to Map GPS Data to Networks Aultman-Hall - August 2001
A Model to Map GPS Data to Networks Aultman-Hall - August 2001
Representation of Networks network accuracy center line representation
Urban Canyons / Moving Vehicle Aultman-Hall - August 2001
Algorithm Development GPS points link route data route choice models traffic line network Optimal GPS Settings PDOP filter noise to signal ratio frequency
Aultman-Hall - August 2001 Lexington, KY Population 250,000 293 Square Miles 1350 miles road
Route Development all road types turns downtown rural and treed areas some aimed to trick do NOT start and end at nodes Aultman-Hall - August 2001
Sample Routes
Sample Routes
Route Dataset 674 routes (18 driven multiple times) average 11.9 miles long maximum PDOP (6,8 & 10) NSR (2,4 & 6) log frequency (1 s, 2 s, 25 ft 50 ft) Aultman-Hall - August 2001
Research Procedure TRIMBLE GPS Receivers Office Pathfinder UNIX ArcInfo - find start and end nodes - use GPS points to adjust link impedances - use MPA to generate route PC ArcView for analysis
Start and End Nodes
Start and End Nodes A B
Start and End Nodes B
Start and End Nodes 35% start and end nodes both right 60% one or other at wrong end of link 2% wrong due to time lag in GPS starting to record points Aultman-Hall - August 2001
Start and End Nodes A B
Start and End Nodes PDOP, NSR and logging did not affect success rate still to come - change buffer size and seek improvement GPS points themselves as stops Aultman-Hall - August 2001
Route Prediction algorithm gets all cases of 16 of 18 of the routes (when start and end nodes correct) approximately 90% correct Aultman-Hall - August 2001
Problematic Route 1
Problematic Route 1
Problematic Route 2
Outstanding Issues start / end missed travel buffer size use inverse of point density as impedance how low can we go on logging frequency
Routing Behavior
Conclusions disaggregate route data collection via GPS is feasible using a MPA to translate the points to network nomenclature is very successful a method to quantify route variation is needed Aultman-Hall - August 2001