Reinforcement Learning-based Box Unloading Sequence Planning for Robotic Container-Unloading System

Abstract

The need for robotic warehouse automation is emerging in the logistics industry because of the risk of manual labor-related employee injuries during truck unloading. The automatic unloading process for boxes in freight containers proceeds in the following order of steps: vision recognition, selection of box to be unloaded, and picking and placing the box on the conveyor belt using a robot manipulator. In this paper, we propose a box unloading sequence plan for robotic-logistics systems. First, it is assumed that the shape of the boxes stacked on the truck is pre-recognized by the visual system. Then, the box unloading sequence plan is generated by reinforcement learning. The performance of the proposed method was verified by comparing it with that of the heuristic method by numerical operation.