Humanoid Robot Imitation Learning Based on COM Correction and Compensation

Since humanoid robot has high degrees of freedom and redundant structure, so the motion planning is complicated when facing different environments. In this paper, kinect was used to collect human motion information as teaching data to conduct the imitation of human motion. The motion planning of humanoid robot was simplified. To maintain the balance of robot in the process of motion, a method was proposed to compensate the robot's center of mass (COM):estimate the deviation of mass center through teaching dat was estimated, and the calculation of angle compensation through Jacobian was conducted and quadratic programming for optimization was optimized. The experiment of imitation system based on robot Nao indicates that the method of COM compensation ensures the balance in the process of imitation, the quadratic programming can effectively ensure the similarity of motion imitation.