Conditionally-minimax filtration in a system with commutating observation channels (Q1285407)

A finite-dimensional nonlinear filtering algorithm is investigated for a stochastic difference observation system in which measurements come from various channels and channel-to-channel switchings are described by a nonstationary Markov chain. The algorithm is based on the conditionally-minimax nonlinear filtering technique. A filter is constructed by properly choosing certain conditionally-minimax nonlinear filtering structural functions inheriting the structure of the mean-square optimal filtering algorithm. The construction of a conditionally-minimax nonlinear filter is described for a nonlinear observation system in which measurements arrive from different channels. The accuracy of the method is experimentally evaluated.