Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : June 28, 2023 - July 01, 2023
In general, a robot dynamics model is derived from the equations of motion, and the dynamics parameters are obtained through identification experiments. However, it is difficult to obtain accurate parameters because of the influence of noise and un-modeled dynamics. These errors may be enhanced by a feedback controller, which degrades control performance. Therefore, it is necessary to identify the un-modeled dynamics from the experimental results. In this study, we focus on periodic motions such as the welding weaving motion by a powered manipulator and propose an identification method of the structure and parameters of un-modeled dynamic. Based on the periodicity of the motion, the quasi-impulse response of the dynamics is derived from the motion error, and its state-space equation is derived. In particular, since the motion errors include both effects of pre-identified parameter error and un-modeled dynamics, they are separated by determining the dimension of the state-space equation. The effectiveness of the proposed method is evaluated by simulations.