Adaptive Penalty Strategies for Electric Vehicle Routing Using Priority-Based Evolutionary Search
Abstract
The rise in the usage of Electric Vehicles (EVs) in logistics has introduced several routing challenges, which mainly include constraints related to battery capacity, vehicle load capacity, and the availability of recharging infrastructure for delivering the goods by the electric vehicle. These factors can make the Electric Vehicle Routing Problem (EVRP) significantly much more complex than classical vehicle routing problems. Most of the conventional evolutionary algorithms often include fixed-weight factors to impose penalties for violating constraints. However, this approach may lead to poor feasibility and slower convergence. This study proposes the usage of various adaptive forms of penalty weight adjustment strategies for solving the Electric Vehicle Routing Problem using Priority-based Travelling Salesman Problem (TSP)-style evolutionary search. This proposed method dynamically adjusts penalty weights based on constraint violation patterns within the population, to improve balance between solution feasibility and constraint satisfaction. The results obtained on the Standard Goeke benchmark instances demonstrate that the proposed adaptive methods can achieve better improvement in the hypervolume results, faster convergence, as well as the feasibility of the solutions obtained.