Anti‐windup compensation for model reference adaptive control schemes

Abstract

Actuator constraints, particularly saturation limits, are an intrinsic and long‐standing problem in the implementation of most control systems. Model reference adaptive control (MRAC) is no exception and it may suffer considerably when actuator saturation is encountered. With this in mind, this paper proposes an anti‐windup strategy for model reference adaptive control schemes subject to actuator saturation. A prominent feature of the proposed compensator is that it has the same architecture as well‐known nonadaptive schemes, namely model recovery anti‐windup, which rely on the assumption that the system model is known accurately. Since, in the adaptive case, the model is largely unknown, the proposed approach uses an “estimate” of the system matrices for the anti‐windup formulation and modifies the adaptation laws that update the controller gains; if the (unknown) ideal control gains are reached, the model recovery anti‐windup formulation is recovered. The main results provide conditions under which, if the ideal control signal eventually lies within the control constraints, then the system states will converge to those of the reference model, that is, the tracking error will converge to zero asymptotically. The article deals with open‐loop stable linear systems and highlights the main challenges involved in the design of anti‐windup compensators for model‐reference adaptive control systems, demonstrating its success via a flight control application.