Deep Reinforcement Learning for Robotic Arm Path Planning in Multi-Obstacle Environments

摘要

This paper presents a novel path planning algorithm for a picking robotic arm in a multi-obstacle environment, based on deep reinforcement learning. The proposed method introduces a new state representation technique that accurately captures the real-time state information of the robotic arm and multiple obstacles using a finite-dimensional representation. This state is represented by the direction vector of the nearest distance to the nearest obstacle around each axis of the robot arm, the real-time angle of the robotic arm, the three-dimensional coordinates of the picking target, and the three-dimensional coordinates of the end effector. The effectiveness of this method is validated through tests in a simulation environment.