Intelligent Control of Manipulator Based on Deep Reinforcement Learning

Abstract

Traditional manipulator control requires the establishment of complex models and mathematical equations, and requires manual adjustment of parameters, which cannot adapt to tasks and environmental changes. With the rapid development of artificial intelligence, deep reinforcement learning algorithms show strong online adaptability and self-learning capabilities for complex systems, and have gradually become one of the important research hotspots in the field of intelligent control and artificial intelligence in recent years. Therefore, this paper combines the deep reinforcement learning algorithm with the manipulator control system, and uses the deep deterministic policy gradient (DDPG) algorithm to achieve the end position control of the manipulator. On this basis, this paper presents a scheme to improve the reward function. Finally, the DDPG algorithm is used for interactive training with the manipulator model. The experimental results show that the deep reinforcement learning algorithm can realize the control of the manipulator, and the improved reward function scheme can greatly improve the stability of training and the task completion rate.