Stable output feedback position control with integral action for robot manipulators

The problem of robust regulation of robot manipulators using only position measurements is addressed. The main idea of the control design methodology is to use an observer for estimating simultaneously the velocity and the modeling error signal induced by model/system mismatches. The controller is obtained by replacing the velocity and the modeling error in an inverse dynamics feedback by their estimates, which leads to a certainty equivalence controller. The resulting controller has a PID-type structure which, under least prior knowledge, reduces to the PI²D regulator. The controller is endowed with a natural antireset windup scheme to cope with control torque saturation. Regarding the closed-loop behavior, it is proven that the region of attraction can be arbitrarily enlarged with high observer gains only, thus semiglobal asymptotic stability can be realized. The result supersedes previous works in the direction of performance estimates; specifically, it is also proven that the performance induced by a saturated inverse dynamics controller can be recovered by the PID-type controller. In this sense, some connections between PID-type and inverse dynamics controllers are revealed.