Multi - Sensor Multi - Message Kalman Filter With Unknown Parameter System

Multi-innovation identification theory is the new midification technology in which the scalar innovation is expanded into the innovation vector, so as to improve the parameters estimation precision of useful information. When the system has untrusted input data, the proposed algorithm can avoid the bad data and the loss of data on the influence of system parameter estimation, and a more accurate parameter estimation is obtained. Multi-innovation identification theory and multi-sensor fusion technology are combined to enhance the system’s fault tolerance and accuracy.Multi-sensor information fusion is to form the characteristic expression of a particular object based on the measurement information from the multisensor. The multisensor information fusion is to obtain the optimal fusion estimation of the system states, so that the best estimation of the state can be obtained, whose accuracy is higher than that of the state estimation based on the single sensor.By applying multi-innovation identification theory and multi-sensor technology for the system with unknown model parameters, the main research results are got and include the following:Firstly,the multi-innovation least squares algorithm and the multi-innovation stochastic gradient algorithm are given for the state space model with unknown model parameters. Based on the transformation between state space model and ARMA model, the state estimation problem is transformed into the signal estimation problem. The estimated parameters of the system is got by using multi-innovation identification method.Secondly,for systems with unknown parameters, a self-tuning multi-innovation Kalman filter is proposed. The estimation of unknown parameters can be achieved by using multi-innovation identification method. The self-tuning multi-innovation Kalman filter is obtained by inserting the valuation of the model parameters into the optimal Kalman filter.Finally,for the multi-sensor system with unknown model parameters, a self-tuning multiple sensor multi-innovation Kalman filter is presented. The estimation of unknown model parameters can be achieved by using multi-innovation identification method. The self-tuning multi-sensor multi-innovation Kalman filters are obtained based on the centralized fusion, the matrix-weighted fusion and CI fusion methods.Some numerical simulation examples verify the effectiveness and correctness of the proposed algorithms.