A unified approach to the decoupled discrete adaptive control of manipulators

The author has previously presented work concerned with the automatic control of industrial-type robots [1, 2]. The quite formidable problems of cross coupling of dynamics, extensive nonlinearities, unknown parameter changes and modelling errors were overcome by a technique of model reference adaptive control. Furthermore, the author has shown [3] that this technique is applicable to most types of industrial robots (according to a classification of Coiffet [4]). In this paper, we show that a simplification of the model of the robot yields a particularly attractive form of model reference adaptive controller, whereby the strategy for each link is decoupled from all the other strategies. The main advantages in this are that the controller design is much simpler than before, and that the amount of computation is reduced. Simulation studies support the view that the loss of information induced by the a priori choice of a simple robot model has little effect on the performance of the closed-loop system.