On stochastic observability of a nonlinear dynamic system (Q1385876)

This is a short note where an extension of earlier results of the author is given. The author considers a nonlinear vector equation in the form \[ \dot x(t)= Ax(t)+ \sum^r_{i=1} g_i h_i(x,t, \alpha_i), \quad t\in [0,T] \tag{1} \] with discrete-time observation \[ y_k= H^Tx_k+ \xi_k, \quad k\in [0,N] \tag{2} \] where \(x\in R^n\) is a vector of state, \(y\in R^1\) is an observation, \(x_k= x(t_k)\), \(y_k= y(t_k)\), \(t_k= \Delta k\), \(\Delta= T/N-1\), \(A\) is an \(n\times n\) matrix, \(H\) is a vector, \(g_i\), \(i=1, \dots,r\) are constant coefficients, \(\alpha_i\) are unknown parameters, \(h_i(x,t, \alpha_i)\) are scalar nonlinear continuous functions, \(\xi_k\) is a correlated second order discrete-time random sequence with zero mean \(E[\xi_k] =0\). The author proposed an algorithm for estimating the vector of state of the system (1) and the parameters \(\alpha_i\).