Decomposed Parameter Identification Approach of Robot Dynamics

A new method is proposed in this paper for the parameter identification of robot dynamics. Different from existing methods, the identification of lOn dynamic parameters for a robot with n Joints is firstly decomposed into n subproblems, each of which deals with the task of identifying 10 parameters. Based on the decomposed models, both sequential off-line and recursive online identification algorithms are then developed. These new algorithms reduce the computational burden greatly and make the parallel computation possible. The determination and processing of unidentifiable and combined identifiable parameters are particularly convenient by using the decomposed models. Algorithm for this purpose is also given in this paper. Finally, simulation results of identifying the dynamic parameters for the first three links of PUMA-560 by using the proposed method are presented to show the effectiveness of this new method.