AN APPROACH TO SUBOPTIMAL FILTERING IN APPLIED PROBLEMS OF INFORMATION-PROCESSING

An approach to suboptimization of the Kalman-Bucy filter is proposed applicable to problems of complex information processing of navigation systems. Suboptimization consists of decomposition of the filter in state space and the time domain. In a region of state space a practically observable subspace of variables is isolated; here a special indicator of estimability of each variable is used. In the time domain transient and stationary modes of filter operation are considered separately. Approximate "standard" analytic expressions are obtained for the amplification coefficient of the filter for a transient regime. This allows them to be implemented quite simply on a ship's navigational calculator. For a stationary mode of operation, adjustment of coefficients is implemented proceeding from minimization of a generalized integral estimate of the quality of transient processes under a given level of variances of output errors from noise disturbances. This increases the margin of stability of the filter and imparts a robustness property. A number of examples are given.