Autonomous Pick-and-Place Using Excavator Based on Deep Reinforcement Learning

Excavator operation demands precise control because of the complicated nature of the tasks. Standard operation of excavator by human worker generally lacks effectiveness and efficiency which is vital in industrial processes. On the other hand, conventional control methods generally rely on the exact model of excavator which is difficult to obtain accurately. In this paper, we proposed an alternative strategy to control excavator using reinforcement learning in which the control policy is discovered through learning process. The Proximal Policy Optimization (PPO) is employed in this research to learn the policy. The results show that the proposed approach can effectively operate the excavator autonomously, especially in controlling the position and orientation of the bucket towards the desired point without prior knowledge of excavator's kinematics. We have successfully implemented the proposed method to perform pick-and-place operation in the simulation scenario.