A State-action Map for Handling Operations with Plural Robots Considering Robots′ Fatigue

Abstract

In this paper, we present an algorithm of task assignment and motion planning of robots used in a palletizing system to pick parts on a conveyor belt. Because of high-speed and continual motion of robots, robot fatigue, which has the potential to bring the system to a halt, becomes a serious problem. To make an optimal algorithm considering robot fatigue and efficiency, we generate a state-action map via reinforcement learning, regarding the whole system an agent. Parts are assigned to the robots based on two simple rules. The robots then chose their optimal motion for pick-up or reassign the parts based on the map. By using the map, robot fatigue is restrained and consequently the efficiency of the system is improved.