Comparative analysis of Direct and Indirect Model Reference Adaptive Control by Extended Kalman Filter

Abstract

Considerations about the increasing complexity of technological systems have stimulated the interest in hybrid systems that can successfully manage switching behaviour or approach nonlinearity. Hybrid systems are made up of two parts: a constant dynamics component and a switching mechanism. This article investigates the effectiveness of direct and indirect model adaptive control approaches for any common tool for hybrid modelling and approximation nonlinear systems. A reference model may be linear or partially refined, specifies the desired loop system behavior that the adaptive controllers are capable of achieving in the face of unknown system dynamics regardless of the system dynamics. Individual control gains are obtained for each subsystem and it is also carefully tuned to the altered behavior of each system. Through the application of dynamic gain adjustment, singularities in the principle of certainty equivalence are avoided indirectly. The state of the reference model is asymptotically monitored for both techniques by assuming that a shared Lyapunov feature is available for the switched reference model.