Adaptive controller design for flexible joint manipulators

Abstract

A new adaptive controller for flexible joint robots is presented here based on singular perturbation theory and using only position and velocity feedback by modelling the 'motor tracking error' Kem as the fast variables instead of the joint elastic forces, where K represents the joint elasticity and em is the motor tracking error. The resulting boundary layer and quasi-steady-state systems are made exponentially stable: therefore the statements of Tychonov's theorem are valid for an infinite time interval.