| Under poor observation conditions,when the measurement equation of the system can’t be verified or calibrated,the use of the measurement equation will often produce unknown system errors,resulting in large filtering errors.Similarly,when the statistical characteristics of state noise and measurement noise are unknown,the performance of the filter will deteriorate,and even cause the filter divergence.The introduction of incremental equation can effectively eliminate the unknown measurement errors of the system.The addition of the noise statistical estimator can also estimate and correct the noise statistical characteristics in real time during the filtering process.Therefore,the state estimation problem of the under-observation system with unknown measurement error and noise statistics can be transformed into the adaptive state estimation problem of the incremental system.To further improve the estimation accuracy of the noise statistical characteristics and state under the under-observation system,this paper focused on multi-sensor under-observation system with unknown measurement errors and unknown noise statistical characteristics.Firstly,a new noise statistical estimation algorithm based on innovation was proposed in this paper.Its estimation accuracy was higher than the existing improved Sage-Husa adaptive algorithm.The proposed adaptive incremental Kalman filter algorithm can effectively solve the problem of unbiased estimation with the statistical characteristics in unknown noises and further state estimation under poor observation conditions.Secondly,In this paper,combined with multi-sensor distributed weighted fusion algorithms,based on the optimal fusion algorithms in the linear minimum variance,the adaptive incremental Kalman estimators of multi-sensor weighted state fusion and weighted observation fusion were proposed respectively.Finally,for the generalized CAR model system,an observation prediction algorithm based on the adaptive incremental Kalman estimator was proposed,and combined with the weighted observation fusion algorithm,which can effectively solve the prediction and update of observation increment under the multi-sensor under-observation system.Simulation examples proved the effectiveness and feasibility of the above-mentioned algorithms. |
