scientific article; zbMATH DE number 3887574

From MaRDI portal
Publication:3222083





The author considers linear stochastic systems with discrete or continuous time-parameter. The systems contain two components, x and y, were x is unobservable. The problem is to estimate x on the basis of observations of y under some restrictions concerning the transition matrix of the system. Section 1 treats discrete-time systems. In Section 2 the processes x and y have the form: \(dx_ t=A_ tx_ tdt+de_ t\), \(dy_ t=C_ tx_ tdt+dv_ t\), \(t\geq 0\), where \(C_ t\) is a given matrix, e and v are random noises with independent increments, and the matrix \(A_ t\) belons to a special class H. The admissible estimator \(\hat x\) for x is of the form of a Kalman-Bucy filter: \(d\hat x_ t=A_ t\hat x_ tdt+K_ t[dy_ t-C_ t\hat x_ tdt]\) where \(K_ t\) is a matrix function. The estimator \(\hat x\) is said to be a mini-max estimator if \(A_ t\) and \(K_ t\) are determined from the condition to realize \[ \min_{K}\max_{A\in H}E[(\hat x_ t-x_ t)^ T(\hat x_ t-x_ t)]. \] The author has found, in explicit form, the solution of this problem in both cases, discrete (Theorem 1) and continuous (Theorem 2).




Cited in
(33)








This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3222083)