Truck Route Choice Modeling using Large Streams of GPS Data

2017Final ReportHunter, BrianPDFTech ReportCAIT-UTC-NC32DTRT13-G-UTC28Decision makingForecastingGlobal Positioning SystemOrigin and destinationPavement management systemsRoute choiceTraffic volumeTravel demandFreight TransportationHighwaysOperations and Traffic ManagementTruck-GPS DataRoute DiversityRoute Choice ModelsChoice Set GenerationUnited StatesFloridaTampa (Florida)Rutgers University. Center for Advanced Infrastructure and TransportationTahlyan, DivyakantPinjari, Abdul R.Luong, TrangOzkul, SeckinUniversity of South Florida. Dept. of Civil and Environmental EngineeringRutgers University. Center for Advanced Infrastructure and TransportationFlorida. Dept. of TransportationUnited States. Federal Highway AdministrationUS Transportation CollectionThe primary goal of this research was to use large streams of truck-GPS data to analyze travel routes (or paths) chosen by freight trucks to travel between different origin and destination (OD) location pairs in metropolitan regions of Florida. Two specific objectives were pursued. The first objective was to measure and analyze the diversity in travel paths chosen by trucks between different OD locations in Florida. Various metrics were used to measure three different dimensions of diversity in truck route choice between any given OD pair. Further, statistical models of diversity metrics were estimated to gain insights on the determinants of various dimensions of truck route choice diversity between an OD pair. The second objective was to evaluate truck route choice set generation algorithms and derive guidance on using these algorithms for effective generation of choice sets for modeling truck route choice. Specifically, route choice sets generated from a breadth first search link elimination (BFS-LE) algorithm were evaluated against observed truck routes derived from large streams of GPS traces of a sizeable truck fleet in the Tampa Bay region of Florida. A systematic evaluation approach based on the algorithm\u2019s ability to generate relevant routes typically considered by travelers and generation of irrelevant (or extraneous) routes seldom chosen is presented. Based on this evaluation, the study offers guidance on effectively using the BFS-LE approach to maximize the generation of relevant truck routes while eliminating irrelevant routes. It was found that carefully-chosen spatial aggregation can reduce the need to generate a large number of routes for each trip. Finally, route choice models were estimated and applied on validation datasets to corroborate the findings