Robot joint friction compensation based on a local modeling technique

This paper considers an approach of local modeling applied to robot joint control to solve the problem of friction compensation. In order to enhance the tracking performance of a robot joint, a mathematical dynamic model is derived from the physical local properties of friction, and is used in the design of a controller to cancel friction-induced errors. The proposed structure does not claim to be faithful to reproduce complex phenomena drown by friction. However, the linearity of the local models simplifies the design and the implementation of the observer and its estimation capabilities are improved with the addition of a nonlinear compensating gain. Experimental tests conducted on a robot joint with high level of friction demonstrate the effectiveness of this observer-based control strategy, for tracking trajectories involving the system in zero velocity region and reversals.