The Algorithm Of Sensor Management Based On Information Theory

The technology of multisensor data fusion uses many sensors to detect targets and obtain the information which from many respects, and it can get better state estimate values than single target. Despite data fusion has such remarkable advantage as redundant and complementary information, but it has much shortcoming as following: uncertain environment, deficient sensor resources and shortage factors of sensor oneself. In order to let the whole system have best capability, foregoing factors lead the unisonous problem of sensor resources distribution to be researched. Besides, along with the people ulteriorly research on data fusion, it is necessary that the problem of sensor management will be researched as a single portion. So, this thesis mainly researches the theories and methods of sensor management based on information theory, and the accomplishment as following: 1. It expatiats the concept of data fusion, and detailedly introducts closed loop system of data fusion, system frame, function, the capability of every module, and the necessary on the problem of sensor management is presented. In view of system, this thesis detailedly researches the theory of sensor management which including concept, the principle of system design, the field of management, function, task and algorithm, and so on. Due to deficient sensor resources, it is necessary that target priority be presented. 2. Basing on statistical characteristic, in order to satisfy the sufficient and necessary condition of resources distribution, this thesis presents a method of sensor management based on combining target priority with information gain. Analyzing the defect of logarithm define on information entropy and the shortage of interacting multiple models (IMM), this thesis gives compare and analysis results by the different defines of entropy and the different filtering models. 3. Basing on movement characteristic, this thesis analyses that the probability density of dynamic target is evolved not only at measurement update but also during measurement. Besides, it detailedly researches the problems of sensor management on dynamic target. 4. In nonlinear system, the problem of optimal estimation lead to the method of sequential Monte Carlo particle filtering is developed, and this method is used to solve the problem of sensor management in nonlinear fusion system.