Genetic algorithm based dynamic scheduling of EV in a demand responsive bus service for first mile transit

Abstract

Demand responsive transit (DRT) services have significantly evolved in the past few years owing to developments in information and communication technologies. Among the many forms of DRT services, demand responsive bus (DRB) services are gaining traction as a complimentary mode to existing public transit services, especially to dynamically bridge the first/last mile connectivity. Simultaneously, the stern regulations imposed by regulators on greenhouse gas emission have enforced electric vehicles (EV) to replace conventional vehicles. However, state-of-the-art (SoA) work proposed to generate routes for EV-based DRB services are inhibited by the low number of ride matches and the excessively high computation time of the algorithms deeming them unsuitable for real-time computations. To this end, we propose a genetic algorithm for dynamic scheduling of EV in a DRB service that reacts to first mile ride requests of passengers. In addition, we also formulate an optimal mixed integer program to generate baseline results. Experiments on an actual road network show that the proposed GA generates significantly accurate results compared to the baseline in real-time. Further, we analyze the benefits of rescheduling passengers and flexible estimated time of arrival of EV to optimize the total travel time of passengers.