No label defined (Q4955384)

The aim of this paper is to show how passivity or dissipativity concepts can be used in the stability analysis of feedback systems incorporating a fuzzy controller. To do this, zero-steady-state-error operations are considered for different classes of two-input-single-output linguistic variable-based fuzzy controllers. Based on a pseudo-time-derivative realization of the linguistic variable `rate of change of error', a SISO continuous-time nonlinear dynamic representation of these fuzzy controllers is proposed. The nonlinear part is shown to be both input and state dependent, with a set of global properties arising from commonly used assumptions on the input and output fuzzy sets. This leads to an output-strong-positive-realness property, which is used in stability analysis. Stability analyses are performed for PI-or PD-like fuzzy controllers and two kinds of plant models are considered, LTI models and 2-degree-of-freedom underactuated Euler-Lagrange systems. For LTI models, stability analysis is done in an absolute stability framework using the Kalman-Yakubovich positive-real lemma. For Euler-Lagrange models, their properties together with those of PD-like fuzzy controlles arising from the product-inference rule are used within the general framework of the feedback interconnection of dissipative systems. For the entire collection see [Zbl 0934.00017].