Hybrid variable neighborhood search for automated warehouse scheduling

We study a new scheduling problem which arise in real-life applications, such as managing complicated warehouses, storage areas, e-commerce malls. Inspired by the automated warehouse of a huge electronic manufacturer, we consider a new picking and packing process on several production lines equipped with parallel machines and intermediate buffer. The picking process is serviced by a limited fleet of transportation robots. Each robot delivers products from the storage to picking stations and back. Moreover, special constraints arise from the availability of parking slots and the duration of the customers’ order handling. For this new makespan minimization problem, we design a hybrid Variable Neighborhood Search(VNS) and Tabu Search(TS) framework. The search for a solution is conducted over a space of order permutations. Original randomized decoding procedure is constructed to evaluate the quality of solutions. Infeasible solutions can arise during the search process, thus we design a special mechanism to return into the feasible domain. We have conducted computational experiments on a set of instances based on real data, provided by the Huawei company with up to 1000 orders, 4 production lines, and 50 robots which corresponds to a typical one-day production plan. The proposed approach provides solutions with average relative error less than 2% from the lower bound.