Random-structure distributed estimation systems (Q1920403)

A linear non-stationary object is considered in discrete time. The observations are linear combinations of state variables distorted by uncorrelated Gaussian errors. During operation, the structure of the measurement system can vary and the structure changes are described by a Markov sequence with given conditional transition probabilities. At each stage, the optimal estimates of the state vector are obtained in the form of conditional means which provides the minimal rms error of MSE. Recurrent Bayesian algorithms for determining the a posteriori probabilities of the structure changes and of the state vector estimation are proposed. As an example a process of radar-measured range of flying vehicles is considered and some numerical results are presented.