Diploma and Master Theses (authored and supervised):
"Do Taxi Drivers Choose the Shortest Routes? A Large-Scale Analysis of Route Choice Behavior of Taxi Drivers using Floating Car Data in Shanghai";
Supervisor: J. Cron, G. Gartner, H. Huang;
Department of Civil, Geo and Environmental Engineering, TU München,
final examination: 2017-09-25.
Understanding taxi drivers' route choice behavior is still a theoretical and empirical challenge. For this end, tradition approaches construct rational choice models and predict deterministically the outcome from a unite set of choices. Although useful, this modeling approach omits a large proportion of elements such as number of left turns, road type, travel distance and travel
time that can in uence taxi drivers' choices in real world. Under this context, the case of Shanghai is of particular interest because it is composed of an enormous amount of taxi drivers' Floating Car Data and includes a large area of traveling activities. In contrast with traditional formal modeling approach, this thesis empirically studies the influence of selected features on the route choice behavior of taxi drivers with a case study in Shanghai. This large-scale analysis uses more than 650 million GPS points and around 3.6 million valid routes from the Floating Car Data. The thesis resolves the questions about whether the shortest routes determine the choice of taxi drivers,and to what extent other potential elements may impact decision making of taxi drivers. Utilizing Floating Car Data in Shanghai, the analysis proceeds first with data pre-processing step which reordered, cleaned and extracted data of interest, as well as using map matching to integrate GPS points to the road network for further analysis. Then an overall pattern of taxi service activities is presented which includes travel areas, frequencies, durations, etc.
Thereafter, two types of analyses are conducted. The first type is comparison analysis which compares the real and the shortest routes in terms of both route-based features and segment-based features. The second type is preference analysis that formulates scenarios and compares different features in each scenario once at a time. While real routes to some extent overlap with the shortest routes with an average value of 46%, the results show that other features, especially the road type, road usage frequency, speed limit and observed travel speed, have impacts on the route choices. In the general situation, the best scenario in combination with travel distance and road usage frequency predicts 53.56% of real routes compare to 46.61% in the
shortest routes. In long distant trips, the scenario with road type and travel distance predicts 51.79% of real routes compared to 13.25% in the shortest routes. What comes as surprise is the insignificance of number of turns in the results in contrast to previous scholarly findings.
taxi driver, route choice behavior, Floating Car Data, Shanghai
Created from the Publication Database of the Vienna University of Technology.