Abstract:
Free-floating car sharing offers a convenient alternative to public transportation in urban regions. However, due to the flexibility of this approach, operators must be able to regularly respond to imbalances in the car sharing system in terms of car location and availability. In order for free- floating systems to operate efficiently, vehicles need to be allocated geographically to meet user demand. Naturally, such systems get imbalanced over time, and the need for relocating vehicles emerges.
This thesis deals with the imbalance problem by presenting a method of data collection and analysis that offers deeper insight into user behavior in free-floating car sharing systems. The method introduced here will contribute to the improvement of operators’ relocation strategies. It does so by answering the question of how vehicles in free-floating car sharing systems should be geographically relocated to best fulfill customers’ needs depending on time of day, weather conditions, and points of interest in a city.
The proposed method represents a behavioral, generic approach to the discovery of knowl- edge about user behavior based on trip data from free-floating car sharing systems. It consists of the collection and normalization of trip data, as well as the subsequent analysis of trip locations, and the application of density-based clustering techniques to find correlations between points of interest and high-demand areas in the car sharing systems. By combining clustering results with geographical points of interest in a geographic information system tool, a wide range of hypotheses can be evaluated. The findings offer recommendations that can potentially improve relocation strategies.
The proposed method is easily applicable for various stakeholders interested in the topic. Us- ing multiple car-sharing providers in Vienna, Austria as a case study, a concrete implementation is presented using open source software. In order to illustrate the validity and applicability of the proposed method, a number of hypotheses are proposed and evaluated. Implications drawn from the results provide three suggestions for the improvement of relocation strategies.
