Observer-based friction compensation in heavy-duty parallel robot control

This paper presents an experimental study on friction compensation for the high-precision tracking control of parallel manipulators. A Luenberger-like observer (LLO) and an extended state observer (ESO) are designed and implemented in real-time control of a 6-DoF heavy-duty Stewart-Gough platform (SGP). The dynamic Lu-Gre model is used in the identification of friction. Performances of the proposed observer-based friction compensators are compared to those of a model-based compensator in computed torque control. Experimental results show that the observer-based compensators significantly improve the tracking performances in high speed motions. Among the investigated observers, the ESO results in minimum RMS error in position tracking. Improvement in position tracking at velocity reversals of the individual leg motions is also observed with the contribution of observer-based compensation. The observer error dynamics is exponentially stable, and the convergence rate can be arbitrarily increased by tuning the observer gain.