Optimization Method of DRL-based Route Planning under Limited Resources

The performance of the route planning module is affected by the on-board computing and storage capabilities, which are growing rapidly but are still limited by the enormous scale of sensed data that need to be stored and processed. So, aiming at the problems of resource-limited deep reinforcement learning-based route planning methods, this paper proposes an optimization method to improve computational effectiveness by using Attention Mechanism and reduce data storage by using improved Double-DQN. The results show that the proposed method enables the planning model to focus on valid sensed information and ignore irrelevant information, as well as effectively reduce the amount of vehicle agent’s experience data storage during environment exploration.