A multi-objective genetic algorithm based approach for dynamical bus vehicles scheduling under traffic congestion

Abstract

Abstract Bus vehicle scheduling is very vital for bus companies to reduce operation cost and guarantee quality of service. Many big cities face the problem of traffic congestion, which leads to the planed vehicle scheduling scheme becoming infeasible. It is significant to study bus vehicle scheduling approaches under uncertain environments, such as traffic congestion. In this paper, a bus vehicle scheduling approach is proposed to handle the traffic congestion. It consists of three phases: firstly, a set of candidate vehicle blocks is generated once traffic congestion happens. Secondly, a non-dominated sorting genetic algorithm is adopted to select a subset of vehicle blocks from the set of candidate blocks to generate a set of Non-dominated solutions. Finally, a departure time adjustment procedure is applied to the Non-dominated solutions to further improve the quality of solutions. Experiments on a real-world bus line show that the proposed approach is able to dynamically generate scheduling schemes and significantly improve the quality of service compared to the comparative approaches.