The Algorithm Of Sensor Management Based On Variable Parameter

The technology of multi-sensor data fusion utilizes many sensors to detect object and obtain the information from many respects, and obtain better state estimate than single sensor. Despite data fusion has some advantages such as redundant and complementary information, it has much disadvantages: uncertain environment, deficient sensor resources and limitary factors of sensor oneself. In order to make the whole system have optimal capability, we need to assign and utilize rationally limited sensors resources. Besides, with further research of data fusion, it is necessary that the problem of sensor management should be researched as a single part. So, this thesis researches the theories and methods on sensor management, and main research work is as following:1. It expatiates on the concept of data fusion, and detailedly introduces advantages of data fusion, development situation and fusion models, and necessity of the problem of sensor management is presented. This thesis detailedly researches the theory of sensor management including concept, the principle of system design, and the field of management, function and algorithms.2. In some algorithms based on Kalman filter and its extension, the presupposition of measurement variance leads to the descent of state estimate accuracy. A self-adaptive estimation algorithm of measurement variance is presented based on spatio-temporal analysis and least squares estimation, which makes full use of redundant and complementary information from the sample data of multi-sensor. Furthermore, when limited sensors track single model objects, in order to assign the sensors resources effectively, this thesis computes the change of information entropy based on the change of error covariance matrix of before and after measurement, and presents an algorithm of sensor management based on self-adaptive estimation of measurement variance by the maximum of information gain on information theory.3. Aiming at tracking of the maneuver objects, in order to accurately depict the moving state of maneuver objects, this thesis adopts interacting multiple models kalman filter. The presupposition of model transition probability leads to descent of filter accuracy. A self-adaptive estimation algorithm of model transition probability is presented which makes full use of the current measurement information, and the thesis presents an algorithm of sensor management based on self-adaptive estimation of model transition probability.4. Using sensor data from different sensors, it can discriminate the belonging type of the object. But the past method is just making using of distance to discriminate the attribution, and it can result in the problem of colossal computing dimension and low efficiency and so on. In order to solve the problems, this thesis presents an algorithm of data fusion based on coefficient of variation.