Optimal Exciting Trajectories for Identifying Dynamic Parameters of Serial Robots

Abstract

Accurate parameter identification of serial robots is of paramount importance to achieve satisfactory performance in performing highly dynamic tasks or obtaining realistic simulations. Focusing on the issue of optimizing excitation trajectory in the parameter identification process, a novel optimization algorithm is proposed by taking into account the interplay among different optimization criteria, in which the initial optimization index is introduced. To this end, the Lie-theory-based dynamic identification approach is introduced to illustrate the impact of trajectory optimization for the identification result, and we conduct a one-way analysis of torque prediction variance to determine whether significant improvements have been made or not. Finally, an experimental comparison between several optimization criteria is implemented for a 7 degrees-of-freedom (DoFs) Franka collaborative robot, the findings show that the proposed approach is capable of minimizing the overall trajectory optimization time while ensuring the identification robustness as compared to several other optimization criteria, which can offer a significant advantage in achieving fast and robust identification.