Evaluating a Simulated Annealing metaheuristic for the Heterogeneous On-Demand Bus Routing Problem with a novel experimental setup

This work addresses the Heterogeneous On-Demand Bus Routing Problem (H-ODBRP), a problem involving the efficient management of transportation requests in an urban setting. Passengers are picked up and dropped off at designated stations near their origin and destination points by a fleet of heterogeneous vehicles. The primary objective is to minimize the total user ride time (URT) for all passengers. A metaheuristic, based on the Simulated Annealing framework, is introduced to provide effective solutions to this problem. A new Integer Programming formulation is also proposed and used to evaluate the solution quality of the proposed metaheuristic on small instances with up to 14 requests. This study contributes to the literature by providing the first insights into the On-Demand Bus Routing Problem using a heterogeneous fleet of vehicles. The results show that, in the instances examined, vehicles with 12 or more seats are rarely utilized. Furthermore, introducing some degree of randomness in selecting vehicle capacity for assigning passengers to empty vehicles proves to be more beneficial than choosing capacity strictly based on demand. A distinctive feature of this study is its focus on the impact of different instance properties, offering a novel experimental setup that could provide fresh insights into on-demand transportation problems.