RP-DQN: An Application of Q-Learning to Vehicle Routing Problems

In this paper we present a new approach to tackle complex routing problems with an improved state representation that utilizes the model complexity better than previous methods. We enable by training from temporal differences. Specifically Q-Learning is employed. show our achieves state-of-the-art performance for autoregressive policies sequentially insert nodes construct solutions on Capacitated Vehicle Routing Problem (CVRP). Additionally, are first Multiple Depot (MDVRP) Reinforcement Learning (RL) and demonstrate problem type greatly benefits over other Machine (ML)