Reinforcement Learning Based Pallet Loading Algorithm and its Application to a Real Manipulator System

Manufacturers pallet loading problem (MPLP) aims to fit the maximum number of boxes into a fixed-size pallet capacity. Solving MPLP can be time-consuming due to its complexity, leading to the use of heuristic methods which may not produce optimal results. This paper proposes a pallet loading algorithm using reinforcement learning to find the optimal solution. Simulation results indicate that the proposed method utilizes the given pallet space more efficiently than the existing heuristic methods. In addition, we introduce a real-life automatic pallet loading system and demonstrate the effectiveness of the proposed algorithm.