Decentralized Model Reference Adaptive Control of Robotic Manipulators

This paper presents a class of decentralized adaptive control schemes based on a model reference approach for a multi-jointed robotic manipulator. An integrated model for the combined dynamics of the manipulator and the joint actuators is developed which provides an ideal framework for the design of decentralized controllers. The adaptive control schemes obtained consist of an on-line recursive estimation of model parameters which are then used to compute the feedback controller gains. To account for the dynamic coupling that exists between the joint motions appropriate feedforward compensation signals are used. Simulation results performed on a VAX 11/780 computer have demonstrated the tracking performance of the adaptive controllers and have illustrated the important connections that exist between the representation of joint motion coupling effects and the controller performance under different tasks and demanded motions. Major strong points of the present schemes are: (i) possibility of implementation by distributed processing techniques and to meet independent joint control objectives, (ii) insensitivity to drift in motor parameters and (iii) possibility of making appropriate selection of control schemes based on specific representations of joint coupling effects and demanded motion characteristics to realize improved performance.