Optimal And Self-calibration Observation Fusion State Valuation And Its Applications

For the multisensor systems, using the weighted least squares (WLS) method, a weighted measurement fusion equation and a equivalent weighted measurement fusion system are obtained. Based on this, using the modern time series analysis method, based on the autoregressive moving average (ARMA) innovation model, two kinds of steady optimal weighted measurement fusion Kalman estimators (filter, predictor and smoother) are presented. Compared with the centralized measurement fusion Kalman estimators, they have the global optimality and completely functional equivalence, and they can reduce the computational burden obviously, so they are adapted for real time applications. For the multisensor systems with unknown noise statistics, the unknown parameters and noise variances of the moving average (MA) innovation model of the subsystem and fusion system can on-line be estimated by two-stage least squares method and the Gevers-Wouters algorithm with dead-band, so that two kinds of self-tuning weighted measurement fusion Kalman estimators are presented. For each kind of them, three kinds of self-tuning weighted measurement fusion algorithms are presented. Their convergences are proved, i.e. if the parameters estimation of the (MA) innovation model are consistent, then they converge to the globally optimal weighted measurement Kalman estimators, so that they have the asymptotic global optimality. The simulated examples for the tracking systems show their effectiveness.