Formulation and Experimental Validation of Two Decentralized Discrete Model-Referenced Adaptive Manipulator Control Strategies

In this paper two decentralized digital model referenced adaptive manipulator trajectory controllers are developed and experimentally tested using a distributed multiprocessor architecture and a PUMA 560 industrial robot. The motivation for decentralized control is the ease and speed of parallel implementation of the developed control algorithms using relatively inexpensive hardware. The first controller is composed of a nominal model-based component that includes nonlinear feedback of local joint variables, and an auxiliary model referenced adaptive component. The purpose of introducing local nonlinear feedback is to compensate, as much as possible, in real time, for manipulator dynamics. The purpose of the model referenced adaptive control signal is to compensate for deviations of the used model parameters from their actual values. Nonlinear interaction and coupling between joints is treated as a disturbance torque that is tuned and compensated for. The second control strategy is purely based on a model referenced adaptive control approach and does not assume any knowledge about the manipulator model. The performances of the two control strategies are compared.