| The main purpose of multi-sensor data fusion is to comprehensively utilize the measurement information of multiple sensors to obtain more accurate state information of the target system.Kalman filtering is the main fusion estimation method for multi-sensor data fusion,but Kalman filtering is only a linear unbiased estimation algorithm,which can only guarantee the minimum mean square error in the linear unbiased estimation class.The biased estimation can obtain a more accurate estimation result under the condition of mean square error by biasing the output result of the unbiased estimation.In addition,Kalman filtering does not consider the constraint information in the system.If the constraint information of the system is integrated into the filtering process,the accuracy of the estimation result can be further improved.In this dissertation,the constrained biased Kalman filtering algorithm is used to further improve the accuracy of multi-sensor data fusion.Firstly,a constrained partial Kalman filter algorithm is proposed,and some properties of the algorithm are analyzed.The existing biased Kalman filter is improved,and a biased Kalman filter whose partial parameters are in matrix form is proposed.In addition,considering that there are usually certain constraints in the actual system,the common equation constraint information is fused into the Kalman filtering process,and the idea of biased estimation is used for reference,and a constrained biased Kalman filter with higher estimation accuracy is obtained.The algorithm is verified by MATLAB experimental analysis.Secondly,the Constrained Partial Kalman Filtering(CBKF)algorithm is combined with the two fusion methods of expanded dimension fusion and sequential fusion,and the sequential fusion algorithm based on CBKF and the expanded dimension fusion algorithm based on CBKF are obtained.The process of these two fusion algorithms is deduced,and the error covariance matrix and the mean square error matrix of the two fusion algorithms are obtained.The accuracy of these two fusion methods is theoretically compared and verified by MATLAB simulation experiments.Finally,a sequential adaptive constrained biased extended Kalman fusion algorithm is proposed.Based on the extended Kalman filter,this algorithm draws on the idea of adaptive filtering in the existing literature,and proposes a new adaptive fading factor.When the model is mismatched,the influence of old data on the filtering results can be reduced,and the filtering divergence can be suppressed,and the simulation verification is carried out through the MATLAB experiment.Research shows,the constrained biased Kalman filtering algorithm proposed in this dissertation can further improve the accuracy of multi-sensor data fusion,and propose a new fading factor,which effectively suppresses the divergence of filtering results when the system modeling is inaccurate.It is of great significance to study multi-sensor data fusion methods and improve the accuracy of multi-sensor data fusion. |
