Dynamic model identification for industrial robots

Abstract

In this paper, a systematic procedure for identifying the dynamics of industrial robots is presented. Since joint friction can be highly nonlinearwith time varying characteristics in the low speed region,a simple and yet effective scheme has been used to identify the boundary velocity that separates this "dynamic" friction region from its static region. The robot's dynamic model is then identified in this static region, where the nonlinnear friction model is reduced to the linear-in-parameter form. To overcome the drawbacks of the least squares estimator, which does not take in any constraints, a nonlinear optimization problem is formulated to guarantee the physical feasibility of the identified parameters. The proposed procedure has been demonstrated on the first four links of the Mitsubishi PA10 manipulator, an improved dynamic model was obtained and the the effectiveness of the proposed identification procedure is demonstrated.