Optimising capacitated EV routing with quantum evolutionary algorithms and federated reinforcement learning

The capacitated electric vehicle routing problem (CEVRP) involves complex interdependencies between dynamic charging and route optimisation. Traditional metaheuristics like genetic algorithms and ant colony optimisation often underperform in large-scale, uncertain EV logistics. This paper presents a quantum-inspired hybrid evolutionary framework (QHIEF) that integrates quantum-inspired evolutionary algorithms (QIEA) with federated reinforcement learning (FRL) for online learning of routing and charging strategies. The problem is modelled as a mixed-integer nonlinear program (MINLP) with stochastic energy consumption and multiple-objective cost minimisation. A refined quantum-inspired differential evolution (QIDE) algorithm enhances route and charging decisions via quantum superposition and adaptive mutation. FRL-based demand forecasting supports real-time vehicle-to-grid (V2G) collaboration. Evaluations on large-scale datasets show a 7.6% reduction in travel distance, 21.3% increase in computational efficiency, and improved convergence over methods like DPCA, CBACO, and SIGALNS, highlighting the framework's potential for scalable, efficient EV logistics.