Discrete-time series identification of sliding dynamic friction in industrial robotic joints

In this paper, the discrete-time series identification of dynamic friction in actuated robotic joints during the sliding regime is proposed. Considering the friction lag as a first order time delay element behind the static friction nonlinearity a reasonable approximation of friction dynamics in sliding is proposed. The regression signal model is derived based on a discrete-time transformation of motion dynamics with nonlinear friction. A robust identification scheme is formulated in the Least-Squares (LS) sense by using an appropriate set of model-related regressors. Further, the related Recursive-Least-Squares (RLS) formulation is provided. The proposed modeling and identification are evaluated experimentally by the offline and online parameter estimation. For these purposes the first vertical rotary joint of the base of a standard industrial robotic manipulator has been used in laboratory environment.