Path Planning Algorithm for Embodied Intelligent Robotic Arms Based on Reinforcement Learning in Human-Robot Scenarios

• 1、School of Intelligent Engineering and Automation, Beijing University of Posts and Telecommunications, Beijing 100876

Abstract：With the rapid development of industrial intelligence and automation technologies, human-robot interaction scenarios are increasingly prevalent in fields such as intelligent manufacturing, medical surgery, and home services. To achieve efficient and safe motion planning, a digital mapping of the human-robot-environment system is designed. Subsequently, a reinforcement learning algorithm based on Proximal Policy Optimization (PPO) is proposed, incorporating techniques such as KL divergence and adaptive clipping mechanisms to enhance the stability and convergence speed of the algorithm, while optimizing its hyperparameters. Finally, the hardware and software environment is configured, and the action space, state space, and reward function are designed. The effectiveness of the PPO algorithm is validated through simulation experiments.

Keywords： Intelligent Robot, Digital Twin, Reinforcement Learning, Path Planning