A novel approach to city bicycle repositioning
3 July 2018
University of Antwerp, Promotiezaal Grauwzusters - Lange Sint-Annastraat 7 - 2000 Antwerpen (route: UAntwerpen, Stadscampus
Prof. dr. Kenneth Sörensen
PhD defence Nicholas Vergeylen - Faculty of Applied Economics
Bicycle sharing has recently become a popular means of public transportation in an urban environment. Users pick up and drop off bicycles at stations that have a limited capacity. Demand and supply of bicycles and empty slots are rarely balanced, which frequently leads to full or empty stations. When a station is full or empty, the users can no longer drop off, respectively pick up bicycles. The repositioning of bicycles by a fleet of vehicles to match demand and supply at the stations leads to complex dynamic problems, such as predicting when a station will be full or empty and how to navigate the repositioning vehicles. These related problems are instantiated, framed, studied and solved.
The contributions of this work are both theoretical and practical. Major theoretical contributions are the proposal and study of a novel routing problem, i.e. the Bicycle Request Scheduling Problem, the idea of using clustering to develop an autotuned GRASP metaheuristic, and the development of an application specific full/empty station predictor. Practical contributions are the development of a methodology to predict full and empty stations, a module that can compute vehicle routes, and the basis of a flexible and automated process of repositioning which still allows for human validation and intervention. The latter is practically desirable.