Research On Route Choice Behavior Of Commuters Based On GPS Data

It is very important to improve the efficiency of overall travel and alleviate the traffic congestion in the city by studying the characteristics of the route selection behavior of the commuter travelers,which can also provide effective technical support for urban traffic planning,management and control.In traditional researches,the data is difficult to be accurately collected.Most of the literatures on the route choice behavior were based on some assumptions,such as:maximum utility theory,prospect theory or regret theory,and then apply it to traffic assignment problem.With the maturation of GPS and GIS technology,the actual route choice behavior of the traveler can be revealed in the high resolution map,and the route choice behavior analysis results will be closer to reality.This paper first establishes a complete method for finding car-based commute travel from original GPS data,and then an off-line map matching algorithm based on multi-path was proposed.The difference between the actual path and the shortest distance path or the shortest time path and the influencing factors were analyzed,and also a comprehensive analysis of the structural reasons of the non-optimal travel was proposed.Then,route choice model for commuters was constructed,and the factors were analyzed.The route choice for multi-day trip between single OD pair was also constructed.Although the research sample of this paper is derived from the survey data of the United States,the research ideas and methods proposed in this paper can solve the same problems in the study of urban development in China,taking into account the commonality of urban development and the portability of research theory and method.The main contents of the paper are summarized as follows:(1)This paper divides the original GPS trajectory into separate GPS tracks that contain only one trip based on the time interval.Consider the equipment error or the investigator's operational errors,the instantaneous speed,average speed or travel time were analyzed to remove the abnormal travel trajectory to get the final effective travel data.Using hierarchical identification method,and based on the starting point and the end of the trip and the type of traveler staff,the car commute travel sample data was found from origin GPS data.And the concept of trip angle was proposed to determine whether there are other purposes on this trip,which might change the chosen route a lot.(2)On the basis of the anomaly data processing,a multi-path map matching algorithm is proposed based on multiple hypothesis techniques was proposed.This algorithm includes six steps:preprocessing,constructing subnetwork,constructing initial path set,constructing path node set and local node set,constructing road segment-node and node-link correlation matrix,constructing potential path set and determining final selection path.Finally,while the actual car commuting route was matched to the electronic map,based on the length of the road or travel time,the shortest distance path and the shortest time path were built respectively.(3)Overlap distribution,relative time difference,absolute time difference and distance difference between the actual path and the shortest distance path or the shortest time path were analyzed.The distribution of distance difference and time difference for different attribute classification were also analyzed.And regression model were employed to analyze the influence of various factors on the deviation between the actual path and the two theoretical paths from the distance ratio,the time ratio and the overlap.Combined with the existing research,for the first time,this article proposed a comprehensive analysis from the selflessness,rationality,perception,information,value,objective,search costs,reliability and pleasure of travel,attempting to understand why travelers do not choose the shortest path.(4)Based on the BFS-LE algorithm,this paper summarized the construction method of the route choice set,and built a new E-BFS-LE for construction of route choice set at the execution level.This paper summarized the discrete model for route choice behavior,and used PS-Logit as the model framework to estimate the parameters collected in this paper and found the influence of different route attributes on route choice behavior.(5)In the sample data including multiday trips,the traveler were divided into SRSP,SRNSP and NSR three categories.At the same time,the average relative time difference between the actual path and the shortest time path is established,including the average overlap between the actual path and the shortest time path,and the standard deviation,were used to describe the characteristics of different types of travelers.And we found that there are different characteristics for different categories.(6)There are significant differences in the route choice behavior for different category travelers.Different from the previous model,in which the travelers are analyzed by a same model,there,the two-stage route choice framework was proposed.And neural network framework and decision tree framework are used to model the two steps respectively.It can be found from the forecast results that this model can describe the behavior of the commuters well.This paper studied the route choice behavior of commuters based on GPS data,and tried to explore the application of GPS and other new technologies in the field of traffic problem research,which also provided a theoretical basis for alleviating urban traffic congestion.