An implementation of a reinforcement learning based algorithm for factory layout planning

Abstract

Factory layout planning is a recurring and time consuming process since multiple often conflicting planning objectives have to be considered simultaneously. Inadequately planned layouts however can significantly impede the operation of a factory. Recent studies from other disciplines have shown the potential of reinforcement learning to solve complex allocation problems. Consequently, this paper presents a reinforcement learning based approach for automated layout planning. In particular, a first implementation of the algorithm using Double Deep Q Learning is presented and used to solve an allocation problem with four functional units and the transportation time as an optimization criterion. The algorithm generated an optimized layout within 8,000 episodes of training and showed promising potential for more comprehensive applications in the future.