Algorithms and Programs of Suboptimal Nonlinear Filtering for Markov Processes

The problem of optimal recurrence estimation of state vector for a nonlinear stochastic dynamic system, which can be multi-modal (multi-structure, switching, logical-dynamical), according to discrete nonlinear measurements is considered. The problem of obtaining real-time estimates on a low power computer is discussed. The covariance Gaussian and Taylorian approximations are used both to the classical infinite-dimensional filter and to a number of new finite-dimensional optimized structure filters. The software package for modeling these filters by the Monte-Carlo method is described.