Micro-Kalman Filtering Fusion For Multi-source Information System With Correlated Noises And Relative Integrated Navigation Algorithms

In the area of information science and technology, the multiple source information fusion is a research topic that has a widespread application backgrounds and important theory significance. In order to improve the precision, timeliness, steadiness and survivability of information processing and the practicability in the harsh environment, the western developed country headed by United States has already widely used the theory and applied research of the multiple source information fusion in varied fields, such as, in the war environment military intelligence, computer networks, communications, the key defense equipment, command integration systems and so on. Although the information fusion research has achieved a great progress in the world in the nearly30years, these conditions of application are restricted. In the statistical decision fusion, the multiple source information needs statistical independence; while in the estimate fusion, the noises measuring in each sensor need mutually independent. In practice, the correlated noises case is a common phenomenon. Hence, this paper is to design a effective data fusion algorithm to solve the problem of correlated noises, and the research work is as follows:Firstly, refer to traditional Kalman filtering fusion formula being in the noise independent of each other restrictive conditions of the forward the problems, we proposed the relevant noise with centralized Kalman filtering information fusion algorithm. The algorithm combines Kalman filtering theory and matrix inverse lemma, the similarity transformation matrix of the technology. And based on noise related centralized Kalman filtering forms equivalent with N noise related micro Kalman filter expression, we proposed a completely noise related multi-sensor system integration of micro-Kalman filter formulas, experiments show that the algorithm not only accurate equivalent to the corresponding centralized Kalman filtering fusion formula, but also can improve the high complexity of the matrix inverse problem.Secondly, refer to traditional Extended Kalman filtering fusion formula being in the noise independent of each other restrictive conditions of the forward the problems, we proposed the relevant noise with centralized Extended Kalman filtering information fusion algorithm. The algorithm based on noise related centralized Extended Kalman filtering forms equivalent with N noise related micro Extended Kalman filter expression, we proposed a completely noise related multi-sensor system integration of micro-Extended Kalman filter formulas, experiments show that the algorithm can improve the high complexity of the matrix inverse problem.Finally, refer to traditional centralized Kalman filtering algorithm in accuracy, robustness, fault tolerance insufficiently, we proposed based on relative combination of federal navigation UKF algorithm. The algorithm firstly through UKF filter with noise-related and federal active filter to calculate the ship navigation and positioning estimates, and then trace the optimal minimum variance criteria matrix trace, and combined with Position estimates for active target ship navigation and relative estimates based on the target ship which gains according to nearby ships weighted fusion which allows for new and more accurate navigation estimates and performance. Experiments show that positioning accuracy of the algorithm is indeed higher than the conventional Kalman filtering algorithm.