Maximum Likelihood Identification of a Dynamic Robot Model

This paper considers the practical implementation of a new maximum likelihood robot identification method, developed by Olsen and Petersen. In particular, the practical issue concerning the estimation of the joint velocities and accelerations from joint angle measurements, and its consequence on the parameter estimation and accuracy, is considered. Simulation and experimental results on a KUKA IR 361 industrial robot are discussed, and compared with models obtained using a much simpler weighted least squares method. KEY WORDS—excitation trajectories, weighted least squares, maximum likelihood, robot dynamics