On the Feasibility of using Neural Networks as High-level solvers for the Pod Allocation Problem within Robotic Mobile Fulfillment Systems

The increasing need for automation in various industrial processes has created opportunities for incorporating robots. In this regard, a Robotic Mobile Fulfillment Sys-tem (RMFS) is a collaborative environment where robots deliver products to humans to fulfill orders. The RMFS is a complex problem that integrates various optimization scenarios. In this work, we analyze the feasibility of using the multilayer perceptron to implement an algorithm selector to improve how we solve the Pod Allocation Problem, one of the challenging subproblems within the RMFS. Our work covers 208 RMFS instances, which are solved with six Pod Allocation Solvers and with the algorithm selectors produced through our approach. Our experiments indicate that this approach leads to competent algorithm selectors. Moreover, such algorithm selectors can rival the best low-level solvers under some particular conditions.