Kinematic Parameter Calibration for 7-DoF Serial Robots Based on Remote Center of Motion and Reinforcement Learning

There is a gap between the actual and nominal kinematic parameters of serial robots, due to errors introduced during production and assembly. However, previous methods usually use laser trackers to achieve closed-loop calibration, which leads to high costs and data dependence. To address this problem, this paper proposes a low-cost kinematic parameter calibration method based on a camera for 7-DoF serial robots. By combining Remote Center of Motion (RCM) for position constraints and Reinforcement Learning (RL) for learning autonomous calibration strategy, we minimize the RCM error to correct the kinematic parameter. Experiments indicated that our method effectively reduces parameter errors by 60.14% and 47% in simulation and the real world respectively.