A new perspective on anti-windup design based on experimental results

Whereas stable linear systems with inputs have a unique steady-state solution, which is independent on the initial conditions and only depends on the input signals, nonlinear systems in general do not possess this property. Due to the lack of this property it is often hard to evaluate the exact behavior of the system, and thus to evaluate the ¿exact¿ performance. In this article we present the results of three experiments performed on a simple nonlinear system, i.e. an anti-windup system consisting of a PI controlled integrator plant with input saturation and linear anti-windup controller. These results, which are validated by simulation and substantiated by theoretical results, clearly show the problems that prevent such an exact performance evaluation of the system. Based on these experimental results we come up with another definition of performance-based anti-windup design, based on the notion of convergent systems.