In this study, the control of picking motion of multiple robot arms with interfering motion ranges for objects flowing on a conveyor by simulation was investigated. The motion trajectories of the robot arms are derived through the implementation of dynamic programming, while reinforcement learning is employed to optimize the picking posture and temporal parameters. The efficacy of the proposed method was substantiated through a comparative analysis of the collection rate with that of the direct control method. In the latter, objects are randomly and continuously flowed, and the timing is completely fixed, in an environment where robot arms may collide with each other under certain conditions.
HIGURASHI et al. (Wed,) studied this question.