Filtering and analysis of linear discrete control systems under a nonparametric specification of the correlated noises (Q1070204)

This work deals with the problem of filtering a linear discrete-time system where the state and observation noises are Gaussian noises which are not necessarily white, but which are specified by there correlation functions. The authors consider the class of Kalman-type filters and derive equations for the gain so that the resulting filter is optimal in this class. This method is then compared with the Bayesian approach and an approximate Kalman filter obtained by extending the state space; the comparison is based on both computational complexity and order of error estimates.