On the stability of the Kalman–Bucy filter with stationary time varying parameters

From MaRDI portal
Publication:4946980





fhe author considers a linear filtering system with stationary ergodic coefficients and continuous time, namely \[ \begin{aligned} dX_t &= (A_tX_t+ F_t) dt+ B_t dW_t+ G_t dY_t,\\ dY_t & = (C_t X_t+ H_t) dt+ dV_t,\end{aligned} \] where \(X_t\) is an unobserved process to be filtered, \(Y_t\) is the observed process, and \((W_t, V_t)\) is the standard Brownian motion. It is proved that if \(X_0\) with given \({\mathcal F}^Y_0\) is Gaussian, then \(X_t\) is conditionally Gaussian given \({\mathcal F}^Y_t\), under the additional assumption that the coefficients of the filtering system are càdlàg functions. It is also established that the Kalman-Bucy filter is exponentially stable under weak conditions on stabilizability and detectability. The author emphasizes that it is important to establish the asymptotic independence of the equations of the filtered system on the initial conditions since these conditions are very often unknown. Two simple examples are presented.











This page was built for publication: On the stability of the Kalman–Bucy filter with stationary time varying parameters

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4946980)