Model reference control and identification of robotic manipulators under uncertainty

Abstract The robotic manipulator is considered in terms of an open kinetic chain with n-degrees of freedom, with nonlinear and nonlinearizable characteristics and couplings, and with uncertain but bounded values of system parameters. The chain is powered by n-actuators and effected by a pay-load which is unknown but within a known band. The objective is to reach for a moving target and avoid moving obstacle. It is specified in terms of a dynamic model to be followed. The well known linear model-reference adaptive control technique is extended to the nonlinear case at hand and used to stabilize the manipulator about the model, at the same time on-line identifying the uncertain parameters and the payload. The method is similar to the self-tuning control which is used in manufacturing when high speed must be combined with high accuracy and not-too-high requirements on sensors (reducing the number of transducers and reading time) as well as broadening the functional abilities of the actuators. The identification uses nonlinear identifier (predictor) system which also stabilizes the resultant response of the manipulator and itself about the desired dynamic model. Algorithms for the adaptive laws in both control and identification are provided. Numerical simulation illustrates the results.