A multi-objective centralised agent-based optimisation approach for vehicle routing problem with unique vehicles

Motivated by heterogeneous service suppliers in crowd shipping routing problems, vehicles’ similarity assumption is questioned in the well-known logistical Vehicle Routing Problems (VRP) by considering different start/end locations, capacities, as well as shifts in the Time Window variant (VRPTW). In order to tackle this problem, a new agent-based metaheuristic architecture is proposed to capture the uniqueness of vehicles by modelling them as agents while governing the search with centralised agent cooperation. This cooperation aims to generate near optimum routes by minimising the number of vehicles used, total travelled distance, and total waiting times. The innovative architecture encapsulates three individual core modules in a flexible metaheuristic implementation. First, the problem is modelled by an agent-based module that includes its components in representing, evaluating, and altering solutions. A second metaheuristic module is then designed and integrated, followed by a multi-objective module introduced to sort solutions generated by the metaheuristic module based on Pareto dominance. Tests on benchmark instances were run, resulting in better waiting times, with an average reduction of 2.21-time units, at the expense of the other objectives. Benchmark instances are modified to tackle the unique vehicle’s problem by randomising locations, capacities, and operating shifts and tested to justify the proposed model’s applicability.