Robotic Mobile Fulfillment Systems: Are Hyper-Heuristics a Feasible Approach When Dealing With the Pod Allocation and Item Storage Problems?

The rise of online shopping has forced warehouses to improve their operations to keep up with demand. In this work, we focus on a warehouse type known as the Robotic Mobile Fulfillment System (RMFS). The literature shows that modern optimization techniques have been explored for the RMFS. However, its complexity restricts the simultaneous optimization of the whole system. For this work, we tackle two poorly explored subproblems: the Item Storage Problem (ISP) and the Pod Allocation Problem (PAP). We analyze the effects of several heuristics (available within the RAWSim-O framework) and hyper-heuristics on the throughput time and the number of orders completed within a 24-hour simulation window. All tests were run in 60 instances, split into four sets. Moreover, we integrated two main stages. First, we considered a baseline given by the best solution for each instance found in a previous work. Our goal was to assess how performance changes when dealing with ISP and PAP simultaneously. We found that although performance improves, no single set of heuristics works best for all instances. The best combination of heuristics provides the best solution in only 23.33% of the instances. Moreover, even if we can solve half the instances with only three combinations of heuristics, we need up to 11 to achieve the best solution in 80% of the instances. In the second stage, we wanted to analyze whether the effect of the ISP extended to sequence-based hyper-heuristics for the PAP. So, we considered a baseline given by the best hyper-heuristic from a previous work. Our data revealed that our best hyper-heuristic improves the solution of 70% of the instances when considering the throughput time, and of 67% of them when considering the number of orders. So, developing a high-level solver that simultaneously selects heuristics for these two problems seems to be a feasible approach to further enhance warehouse performance.