Confidence set-membership FIR filter for discrete time-variant systems (Q6073030): Difference between revisions

The authors deal with the filtering problem for stochastic systems described by the discrete-time equations \[ \left\{ \begin{array}{ll} x_{n+1} = A_nx_n + B_nu_n + E_nv_n, \\ y_n = C_nx_n + D_nu_n + F_nv_n, \end{array}\right. \] where the vectors \(x_n\in\mathbb R^{dx} , u_n\in\mathbb R^{du}\) and \(y_n\in\mathbb R^{dy}\) refer to the state, input and measurement, respectively. \(A_n, B_n, C_n, D_n, E_n, F_n\) are system matrices. The noise vector \(v_n\in\mathbb R^{dv}\) consists of two mutually independent unknown but bounded noise and the Gaussian noise (come from separate sources, for example). A confidence set-membership finite impulse response (FIR) filtering problem for the confidence level \(1-\alpha\) (\(\alpha\) is the significance level) consists in construction of a set of state estimates (confidence state set) \(\widehat{X}_n(\alpha)\) such that the probability measure \(P\{x_n\in \widehat{X}_n(\alpha) \}\geq 1-\alpha \) and the size of \(\widehat{X}_n(\alpha)\) is minimal. The authors proposed a CSM-FIR filter algorithm that provide confidence state set that includes the true state with an adjustable confidence level. An analytical solution and an LMI-based numerical solution for the filter gain are derived by minimizing the size of the confidence state set. An example is provided to illustrate the effectiveness and superiority of the proposed filtering algorithms.