Deep Reinforcement Learning for the Electric Vehicle Routing Problem With Time Windows

The past decade has seen a rapid penetration of electric vehicles (EVs) as more and logistics transportation companies start to deploy for service provision. In order model the operations commercial EV fleet, we utilize routing problem with time windows (EVRPTW). this paper, propose an end-to-end deep reinforcement learning framework solve EVRPTW. particular, develop attention incorporating pointer network graph embedding layer parameterize stochastic policy solving is then trained using gradient rollout baseline. Our numerical studies show that proposed able efficiently EVRPTW instances large sizes are not solvable current existing approaches.