Abstract
This paper studies an autonomous order picking system with two types of robots. To implement a retrieval transaction, the retrieval robot first removes a bin to the buffer position located at the first tier of the system. Then, a transport robot moves the bin to a station for order picking. We investigate the optimisation problem of order sequencing and assigning bins and buffer positions to orders, to minimise the total number of bin moves. We construct an integer linear programming model and design an improved meta-heuristic algorithm to solve the model. Several crossover and mutation operators are introduced to perturb the order picking sequence, and two insertion operators are developed for goods sequencing. We validate the algorithm on various instances and conduct numerical experiments to explore the effect of key parameters on the order picking efficiency. We find that the performance of the algorithm is insensitive to the number of orders, especially when the system has a small buffer capacity and few types of goods. Deploying a large buffer area will improve the picking efficiency, but unlimited increases in buffer capacity will harm the marginal revenue provided by the buffer. Additionally, buffers close to pick stations demonstrate greater benefits.