Identification and control of electro-mechanical systems using state-dependent parameter estimation

This paper addresses the important topic of electromechanical systems identification with an application in robotics. The standard IDIM-LS method of identifying models for robotic systems is based on the use of a continuous-time inverse dynamic model whose parameters are identified from experimental data by linear Least Squares estimation. The paper describes a new alternative but related approach that exploits the State-Dependent-Parameter (SDP) method of nonlinear model estimation and compares its performance with that of IDIM-LS. The SDP method is a two-stage identification procedure able to identify the presence and graphical shape of nonlinearities in dynamic system models with a minimum of a priori assumptions. The performance of the SDP method is evaluated on two electromechanical systems: the Electro-Mechanical Positioning System (EMPS) and the second link of the TX40 robot. The experimental results demonstrate how SDP identification helps to avoid over-reliance on prior conceptions about the nature of the nonlinear characteristics and correct any deficiencies in this regard. Finally, a simulation study shows how the resulting SDP model is able to facilitate nonlinear control system design using linear-like design procedures