A two-stage decentralized adaptive controller for robotic manipulators

This paper presents a two-stage adaptive control strategy for robotic manipulators by using decentralized control technique. In the first stage, which is performed off-line as the feedforward, the nominal torques are computed from the available dynamic equations. The second stage is performed on-line as the feedback, in which, the modification torques are generated to drive the trajectory deviations to zero. This feedback stage is designed based on a perturbation model expressed in the decentralized form, in which every joint subsystem is given by a two-dimension state equation. The modification torque of every joint is determined by minimizing a generalized cost function involving the weighted position error, velocity error and modification torque of the joint. As the adaptive controllers of all the joint subsystems are independent of one another, multiple CPU´s are used for parallel processing. Simulations to a manipulator with three degrees of freedom under strong load disturbance are presented to demonstrate the effectiveness of this control strategy.