Research on Motion Stability Control Algorithm of Multi-axis Industrial Robot Based on Deep Reinforcement Learning

Research on Motion Stability Control Algorithm of Multi-axis Industrial Robot Based on Deep Reinforcement Learning

연구 분야: Infrastructure

학회: International Conference on Multimedia Technology and Enhanced Learning

When the multi-axis industrial robot is disturbed by the outside, there will be some problems, such as large deviation of heading angle, time-consuming motion control and long one-way adjustment distance. Therefore, a motion stability control algorithm of multi-axis industrial robot based on deep reinforcement learning is proposed. Design the structural characteristics of multi-axis industrial robot, consider that the foot position can reach the expected motion range, and optimize the motion trajectory planning mode of multi-axis industrial robot. According to the expected movement speed and angular velocity of the center of mass specified artificially in advance, the position vector of the nominal foothold of the foot end of the multi-axis industrial robot is calculated. Based on deep reinforcement learning, the dynamic equation is constructed, and the motion stability control algorithm is designed by describing the change rate of the momentum of the center of mass. Experimental results show that the average deviation of heading angle of the proposed algorithm is 5.92, the average time of motion control is 2.57 s, and the average distance of one-way adjustment is 31.78 mm, which shows that the proposed method has good fault tolerance of motion stability control and can effectively improve the accuracy and efficiency of motion stability control.