A deep reinforcement learning hyper-heuristic to solve order batching problem with mobile robots

In e-commerce logistics, it is critical to enhance the efficiency of the order-picking system. Motivated by applications of automatic logistics, we consider the mobile robot based order batching problem. In this problem, mobile robots carry shelves to the picking station for order picking and then return them. The objective is to reduce shelf movements while minimizing the number of delayed orders. We introduce a hyper-heuristic method based on deep reinforcement learning to optimize the order batching strategy in the system. The proposed method adaptively selects the order batching strategy, significantly improving the sequential decision-making process in order picking. Through extensive tests, we demonstrate the superiority of the proposed method over several existing heuristic methods in a range of test scenarios. The results show that the proposed method outperforms other existing heuristic methods in a range of test scenarios, offering more stable and effective solutions. This study is a pioneer in the application of deep reinforcement learning to the mobile robot based order batching problem, offering a novel perspective and methodology to overcome the challenges of sequential decision-making optimization in order picking systems.