| Summary: | Vehicle scheduling problem (VSP) in public transit refers to arranging a fleet of vehicles to make vehicles’ departure times coincide with the times in given bus timetables. It is vital for bus enterprises to ensure the service quality and save operational cost. Due to the limited driving range and charging requirements of electric vehicles, how to efficiently schedule electric vehicles is a challenging task with the popularization of electric bus vehicles. In this paper, we propose a two-stage solution approach (SA-LS) combining a simulated annealing (SA) and a local search (LS) for an electric vehicle scheduling problem (EVSP) in public transport. This approach is based on a novel solution coding that represents a scheduling solution by departure times of a fleet of vehicles. The decoding procedure considers limited driving range and recharging time of electric vehicles. The SA is used to find a set of solutions able to be potentially improved by LS. The LS combined with a departure-time adjustment procedure (DTAP) is devised to improve the solutions found by SA. Evaluation functions are devised separately for SA and LS to guide their search. The proposed approach is applied to real world vehicle scheduling problem of three bus lines in Qingdao city, China. Experiments show that SA-LS is able to generate high-quality scheduling solutions within short computational time.
|
|---|