Robust feedforward control of robotic arms with friction model uncertainty

To design feedforward controllers for robots, a model that includes friction is important. However, friction is hard to identify, which causes uncertainty in the model. In this paper we consider rest-to-rest motions of robotic arms that use only feedforward control. We show that it is possible to design feedforward controllers such that the final position of the motion is robust to uncertainty in the friction model. We studied a one DOF robotic arm in the horizontal plane, of which we show analytical, simulation and hardware results and we also show simulation results of a planar two DOF arm. Our friction model includes three types of friction: viscous, Coulomb and torque dependent friction. The results show that it is possible to eliminate the sensitivity of the final state to uncertainty in the three types of friction.