Multi-sensor Centralized And Distributed Information Fusion Filter

Multisensor information fusion is also called multisensor data fusion or multisources information fusion. More accurate and perfect conclusions can be gained for multisensor systems oriented to complex applications by dealing with multi-stage and multi-level multisensor's information.Multisensor information fusion estimation is one of the important branches in multisensor information fusion technology.By using the detected data for the same target, based on some optimal fusion rules, the optimal fusion estimation value can be obtained creditably and accurately. And the accuracy is higher than local estimations.Self-tuning information fusion filtering is a frontier field between optimal information fusion filter and system identification,It can deal with information fusion filter and signal estimation problems for the multisenor system with the unkonw model parameters and noise statistics.So it has important theoretical and engineering value.For the multisensor systems with uncorrelated noises and unknown noise variances, substituting the information fusion noise variance which based on the correlative method into the centralized fusion optimal information filter, a self-turning centralized fusion information filter is presented. Comparing with the self-turning centralized fusion Kalman filter based on the Riccati equation, it can avoid computing the inverse matrix with high dimension, so that the computational burden can be reduced significantly.Substituting the fusion noise variance estimator into the optimal distributed fusion information filter, a self-tuning distributed fusion information filter can be obtained. Comparing with the self-turning distributed fusion Kalman filter, it can be easy to diagnose and separate the malfunction.For the multisensor block-companion system with unkonw parameters and unkonw noise variance, using the recursive instrumental variable (RTV) method and correlative method, the estimations of the unkonw parameters and noise variance are obtained. Substituting the estimation into the optimal distributed fusion information filter, a self-tuning distributed fusion information filter can be obtained.By applying the dynamic error-system analysis (DESA) method, it is shown that the self-tuning centralized and distributed information fusion filters introduced above are converged to the optimal centralized and distributed fusion information filter with probability 1, so that they have asymptotically global optimality.