A unified framework for dynamic parameter identification of elastic joint robots with hysteresis and friction nonlinearities

The accuracy of dynamic parameter identification significantly affects the model-based control performance of elastic joint robots (EJRs). Currently, the modeling and identification of friction and hysteresis nonlinearities in EJRs are mostly treated as separate steps requiring specialized identification equipment. This complicates the identification process and makes parameter re-identification more difficult. Moreover, some parameter identification results could be invalid either because they violate physical constraints or contribute minimally to dynamics. To address these challenges, this paper proposes a unified dynamic parameter identification framework for EJRs that integrates nonlinear friction and hysteresis models without requiring additional processes or equipment. The framework introduces physical feasibility constraints and iterative reweighted least squares to obtain physically feasible dynamic parameter solutions and deal with heteroscedasticity. It also includes a parameter update strategy that addresses parameters with minimal contribution to dynamics. Verification experiments conducted on a 2-degree-of-freedom EJR validate that the proposed framework is reliable and accurate. To the best of our knowledge, this is the first physically constrained dynamic parameter identification framework for EJRs that integrates nonlinear friction and hysteresis models.