Stability analysis and nonlinear observer design using Takagi-Sugeno fuzzy models. (Q607638)

Many problems in decision making, monitoring, fault decision and control require the knowledge of observers of state variables and time-varying parameters which are not directly measured by sensors. For linear models a wide array of estimation techniques are available, such as variants of the Kalman filter. But no general methods exist for the design of estimators for nonlinear systems. The design and tuning in nonlinear case is generally complicated and involves long computational costs. The reviewed monograph provides a range of methods and tools for design observers for nonlinear systems represented by a special type of the dynamic nonlinear Takagi-Sugeno (TS) fuzzy model, which is a convex combination of affine linear models and is known to be a universal approximation for a broad class of nonlinear systems. The indicated structure of TS models facilitates stability analysis and observer design by using effective algorithms based on Lyapunov functions and linear matrix inequalities. Dynamical systems are modeled here in the state space framework, using a state transition model, which describes its evolution in time and a measurement model, which relates the measurement to the states, accordingly to the system \[ \dot{\mathbf{x}}(t)=f(\mathbf{x}(t),\mathbf{u}(t),\theta(t)),\;\;\; \mathbf{y}(t)=h(\mathbf{x}(t),\mathbf{u}(t),\zeta(t)), \tag{1} \] where \(\mathbf{x}\) is the vector of the state variables, \(\mathbf{u}\) is the vector of the input or control variables, \(\theta\) and \(\zeta\) are (unknown/uncertain) parameters and denotes the measurement vector. In order to design observers this nonlinear system is represented by TS fuzzy models of the form \[ \dot{\mathbf{x}}(t)=\sum_{i=1}^m w_i(\mathbf{z}(t))(A_i\mathbf{x}(t)+B_i\mathbf{u}(t)+a_i),\; \;\; \mathbf{y}(t)=\sum_{i=1}^m w_i(\mathbf{z}(t))(C_i\mathbf{x}(t)+c_i), \tag{2} \] where \(m\) is the number of local models, \(A_i,B_i,C_i\) are the matrices and \(a_i\) and \(c_i\) are the biases of the \(i\)-th local model, \(\mathbf{z}(t)\) is the vector of the scheduling variables, which may depend on the states, inputs, measurements or other exogenous variables, and \(w_i(\mathbf{z}(t))\) , \(i=\overline{1,m}\) are normalized membership functions, i.e., \(w_i(\mathbf{z}(t))\geq 0\), \(\sum_{i=1}^m w_i(\mathbf{z}(t))=1, \forall t\). Chapter~2 contains the necessary notions and background, in particular the dynamic TS-system together with two methods for construction of TS models. The first one, the sector nonlinearity approach, is used to obtain the exact fuzzy representation of the nonlinear system in a compact set of the state-space, while the second one gives the Taylor series expansion in several operating points to obtain an approximate model. The aim of Chapter~3 is to review various results concerning the stability analysis and control design of TS-models at the usage of Linear Matrix Inequalities (LMI) technique and Lyapunov functions theory. Among the numerous possible choices of results such as state and output feedbacks with uncertainties, \(H_2,H_{\infty}\) performance delays, the authors present here works to give the main ideas of various possibilities of TS-models. In particular, for state feedback the results for performance through attenuation, and robust control of TS-models with uncertainties are given and also the Input-to-State Stability (ISS) property for exogenous signals is investigated, illustrated on several examples. Chapter~4 introduces the concept of observers used for TS fuzzy systems and reviews methods for designing observers such that the estimation error asymptotically converges to zero. Here the observer-based stabilization is described as the extension of Chapter~3 methods. In Chapter~5, techniques for the stability analysis and observer design of cascaded systems (as important class of distributed systems) are presented. First, an algorithm to participation a general nonlinear system into a cascade of two systems is presented, together with stability conditions for cascaded nonlinear systems. In the second part the stability analysis of cascaded TS-systems is given. The third one studies cascaded observer design for TS systems. In Chapter~6, the large-scale systems as a class of general distributed systems are considered, which are not cascaded but distributed. It means that the influence among the subsystems is not in only direction. Here each subsystem is represented by a TS fuzzy model, but the subsystems are coupled. Methods for the distributed stability analysis and observer design for such TS fuzzy systems are presented on the base of common quadratic Lyapunov functions and easy solvable LMI conditions. Such approach gives several advantages compared with the centralized approach, among which flexibility and reduced computational costs. In Chapter~7 the TS systems which are changing over time or are influenced by unknown inputs are considered for which adaptive observers can be designed. Such observers simultaneously estimate the states and unknown inputs or parameters of a system. The design of observers in the presence of unknown inputs is a real important problem, since very often not all inputs are known, and the unknown inputs may represent effects of actuator or plant component failures. The observers designed on the base of known part of the fuzzy model guarantee either the asymptotic convergence to zero or the boundedness of the estimation error.