Design of Dynamic Obstacle Avoidance and Adaptive Trajectory Tracking Algorithms for Industrial Robots Enabled by Reinforcement Learning

To address the issues of insufficient obstacle avoidance safety, low trajectory tracking accuracy, and poor robustness of industrial robots in dynamic and complex scenarios, this paper proposes a fusion scheme for dynamic obstacle avoidance and adaptive trajectory tracking of industrial robots based on the improved proximal policy optimization (PPO) algorithm. Firstly, a reinforcement learning state space and action space integrating the robot's posture, obstacle positions, and motion states are constructed, and a multi-objective reward function is designed to balance the obstacle avoidance safety and trajectory tracking efficiency. Secondly, a dynamic obstacle avoidance model is built based on the improved PPO algorithm, and an adaptive controller that optimizes PID parameters through reinforcement learning is combined to achieve collaborative control of the obstacle avoidance strategy and trajectory tracking. An experimental environment is built on the ROS and Gazebo platforms. Experimental results show that in industrial scenarios with dynamic obstacles moving at a speed of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\leq 0.8 ~\mathrm{m} / \mathrm{s}$</tex>, the obstacle avoidance success rate reaches 98.7 %, the trajectory tracking error is reduced by 32.4 % compared to the traditional model predictive control algorithm, the response time is shortened by <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{2 8. 1 \%}$</tex>, and it still maintains good robustness under environmental disturbances. This algorithm can effectively improve the motion control performance of industrial robots in intelligent manufacturing scenarios and provide technical support for the execution of dynamic and complex tasks of industrial robots.