Using GPS data in route choice analysis : case study in Boston

Abstract

The pervasive location-based technologies, such as GPS and cell phone, help us find the pattern of geographical information of human behavior and also help dig opportunities in real world. In transportation field, they help people better understand the transportation behavior and at the same time collect necessary information for us. One important aspect of its application is how people choose the route given the existing urban network. However, dealing with the excessive amount of data and the modeling of route choice behavior are two major challenges in the route choice analysis.This thesis discusses the general process in the route choice analysis, from GPS data processing, map matching to the generation of route choice sets. Besides, the Path-Size logit model is implemented to address the modeling issue. In this thesis, I develop a new effective method, which I called Point-Based Local Search Map Matching, to match the consecutive GPS data to the network data. Also, I develop a new model, which I called Random Weight Choice Set Generation Model to deal with the choice set generation problem in the route choice analysis. The data comes from two major sources. One is the Boston car GPS data. It tells when and where a specific car is. The other is the Boston urban network data, which contains all types of roads in GIS format.