Optimization Control Method of Robot Arm Based on Multi-Agent Reinforcement Learning

Due to the rapid development of industrial automation and intelligent manufacturing, robotic arms have been widely used in many fields. This study analyzes the limitations of traditional control methods in robot arm control and introduces the concept of multi-agent reinforcement learning (MARL), which treats each joint or functional module of the robot arm as an independent agent that collaborates to complete complex tasks. The research results indicate that, compared with traditional recurrent neural network (RNN) and deep neural network (DNN) control methods, the method based on MARL exhibits significant advantages in multitask control. In terms of task completion rate, this method has improved by an average of about 10%, especially in complex tasks where it performs even better. This method also effectively reduces the fluctuation of task completion rate, demonstrating better stability and consistency. The method based on MARL has significant advantages in promoting effective collaboration among agents and improving task execution efficiency.