Trajectory Planning Algorithm of Manipulator in Small Space Based on Reinforcement Learning

The development of reinforcement learning has driven the progress of robot control technology. In recent years, reinforcement learning has become one of the highly concerned fields in the academic community, especially the control of robotic arms in the industrial field. In order to achieve intelligent and efficient production, the emphasis is on the research of obstacle avoidance motion planning of the manipulator. However, traditional trajectory planning algorithms have problems such as slow convergence speed, low intelligence, and difficulty in achieving optimization. In this regard, this research takes the six degrees of freedom manipulator PUMA550 as the research object, and focuses on the obstacle avoidance motion planning problem of the manipulator, studies the manipulator modeling based on the improved D-H parameter method, Rapidly-exploring Random Trees (RRT) algorithm, the Q-learning algorithm and the double Q network learning alzorithm.