Adaptive PD control of flexible joint robots without using the high-order regressor

This paper address adaptive control of flexible joint robots with possible uncertainties in both the manipulator and actuator dynamics. Most of the previous works require the computation of a second-order regressor plus acceleration or even jerk feedbacks when the joint stiffness is weak (or elasticity is strong). The adaptive controller proposed here does not need the second-order regressor while still possessing proven asymptotical stability. It also avoids the acceleration feedback by means of an observer, which is independent of robot parameters. The coupling effect between the tracking errors and the observation error is explicitly studied in the closed-loop stability analysis. As the result, the proposed adaptive controller has almost the same complexity as the adaptive controllers for rigid-body robots