State reachability analysis of fuzzy models (Q1925053)

The paper discusses a problem of reachability analysis in fuzzy models described in terms of a series of ``if-then'' rules assuming the form \hskip17mm if \(x(k)\) is \(C_1\) then \(x(k+1)=A_1x(k)+B_1 u(k)\) \hskip17mm .... \hskip17mm if \(x(k)\) is \(C_p\) then \(x(k+1)=A_px(k)+B_p u(k)\), where \(x(k)\in\mathbb{R}^n\), \(u(k)\in\mathbb{R}^m\) and \(C_i\), \(i=1,2,\dots,p\), are fuzzy relations defined in \(\mathbb{R}^n\) while \(A_i\) and \(B_i\) are constant matrices. The state \(x(k_f)\) is reachable from \(x(0)\) if there is a sequence of controls \(u(0),u(1),\dots,u(k_f-1)\) for which the system proceeds from \(x(0)\) to \(x(k_f)\) (\(k_f\) is fixed). The reachability problem is solved under auxiliary assumptions about the model, namely (i) all membership functions are Gaussian-like, (ii) the decoding (defuzzification) is realized with the centre of area method, and (iii) the composition of the rules exploits the product operator. It is shown that the problem can be converted into a sequence of tasks of nonlinear programming. The analysis of computational complexity is also included.