State smoothing and filtering using a constant memory for dynamic systems with nonlinear interference and missing obvservations (Q916628): Difference between revisions

The estimation of the state of a partially observed nonlinear discrete-time stochastic system by finite-state approximation is proposed. Previous schemes require an exponentially increasing memory with time. Here the state estimation is carried out in blocks so that its implementation requires only a constant memory. Numerical simulations are presented. No convergence property is discussed.