A new approach to the design of a fuzzy sliding mode controller. (Q1408839)

The paper is concerned with the development of a generalized fuzzy sliding mode controller. The model under discussion is governed by the expression \[ d^n y/dt^n= f(y, dy/dt,\dots, d^{n-1} y/dt^{n-1}, t)+ u \] where ``\(f\)'' is composed of its nominal (known) part \(f\) and an uncertain (unknown) part \(\Delta f\) which is bounded by some known function \(F\), that is \(|\Delta f({\mathbf x},t)|\leq F({\mathbf x},t)\) where \[ {\mathbf x}= [y,dy/dt\cdots d^{n-1}y/dt^{n-1}]^T. \] The control objective is to determine a continuous sliding control \(u\) such that the output ``\(y(t)\)'' tracks a target \(y_d(t)\). The proposed generalized error transformation reads in the form \[ s_c= \Biggl({d\over dt}+\lambda\Biggr)^{n-1} \Biggl({d\over dt}-\lambda\Biggr) w(t) \] with \(\lambda> 0\), \(e(t)= y(t)- y_d(t)\) and \(w(t)= \int^t_0 e(\tau)d\tau\). The proposed fuzzy controller is composed of rules of the form ``if \(s\) is \(A\) and \(s_c\) is \(B\) then \(u= C\) with \(A\), \(B\), and \(C\) being fuzzy sets of \(s\), \(s_c\), and \(u\), respectively. The application of this controller to robotics is presented as well.