Stability and performance analysis for non-model-based friction estimators

Many motion control systems use friction cancellation to improve precision. In nonmodel-based methods, rigid body dynamics and motion measurements are used to estimate the unknown friction force acting on the system, without posing a structured nonlinear friction model. In this paper, the stability and experimental performance of three non nonmodel-based friction estimators are considered. These include a Kalman filter based approach, a predictive filter based approach and a local function estimation approach. Eigen analysis is used to determine the stability margins for the first two approaches, and Lyapunov analysis is used for the third. The analyses show that all three friction compensators are stable for a broad range of operating conditions and that experimental observations match theoretical predictions of stability margins. Experimental results show that the Kalman filter performs best overall, while the difficulty in tuning constrains local function estimation. The predictive filter friction estimate has undesirable high frequency content which when filtered leads to improved performance