Unknown parameter identification of parameterized system using multi-layered neural network

Parameter identification of dynamical systems

“…This feature renders in the design of a new neural network different from commonly used networks based on the delta-learning rule. It has been established that the identification of the unknown parameters does not depend either on the stability of the underlying dynamical system or on the stability of the auxiliary polynomial introduced in this method, unlike as in [4]. Interestingly enough, the conclusions of this work are based on the elegant simulation results presented in this paper, but these results call for an analytical approach and this problem will be addressed in a subsequent exposition.…”

“…(2), unlike as in the case of Kim and Lee [4], proceeding as in [4], the reparameterized form of the plant may be written as…”

“…Yadaiah et al [12] have obtained the parameters of state variable model for a dynamical system with knowledge of inputs, states of the system and derivatives of the state variables. Kim and Lee [4] have attempted this problem in which multi-layered neural networks are employed to estimate parameters of input-output model of a linear dynamical system under the assumption that the input-output functions are piecewise continuous and bounded on R þ ¼ ½0; 1Þ. In real world problems it is very difficult to accomplish that the input-output functions of any dynamical system are continuous and smooth.…”

“…Kim and Lee [4] have introduced an auxiliary polynomial and required that this polynomial should be Hurwitz, which means that, all the characteristic roots of this polynomial have negative real parts. Our work also requires the use of a polynomial and we establish that for the purpose of tracking the unknown parameters, the polynomial need not be of Hurwitz type.…”