Deterministic Annealing for Depot Optimization: Applications to the Dial-A-Ride Problem

This paper introduces a novel meta-heuristic approach to optimize depot locations for multi-vehicle shared mobility systems. Dial-a-ride problem (DARP) is considered as a case study here, in which routing and scheduling for doorto-door passenger transportation are performed while satisfying several constraints related to user convenience. Existing literature has not addressed the fundamental problem of depot location optimization for DARP, which can reduce cost, and in turn promote the use of shared mobility services to minimize carbon footprint. Thus, there is a great need for fleet management systems to employ a multi-depot vehicle dispatch mechanism with intrinsic depot location optimization. In this work, we propose a deterministic annealing meta-heuristic to optimize depot locations for the dial-a-ride problem. Numerical experiments are conducted on several DARP benchmark instances from the literature, which can be categorized as small, medium and large based on their problem size. For all tested instances, the proposed algorithm attains solutions with travel cost better than that of the best-known solutions. It is also observed that the travel cost is reduced up to 6.13% when compared to the best-known solutions.