A Study Of Sensor Management

With the rapid development of science and technology, especially for sensor, signal detection and processing, and computer, multisensor data fusion technique is also developing rapidly. At the same time, it is a great need especially for modem weapon systems. In order to make the best use of multisensor fusion system, the sensor management is put forward and becomes an important part of data fusion system. Here the theory and method of sensor management are studied systematacially and the main contributions are as follows: I. From the viewpoint of system, the closed loop control scheme of data fusion systems is proposed which is composed of sensor, data fusion, decision making and sensor management subsystems. Then, the probability information model of data fusion and information filters of state estimation and discrete state classification are given respectively for different structures and algorithms. On this basis, the basic frame of sensor management is put forward. 2. The idea of dynamic programming is introduced into sensor management. When the detection and false alarm probability distributions are restricted by a symmetry condition, The optimal strategy will only depend on the ordering of the conditional probability of the hypotheses. For target classification, the constraints of sensor resource may be transformed into the expected constraints, and the multitarget dynamic programming can be decoupled into single target one. 3. The finite set theory is introduced into sensor management, and the performance measure of single-sensor single-target system can be extended to the multisensor-multitarget ones. 4. On the above basis, a set of feasible methods of sensor management is presented for the different cases. ? Based on the quantitative formulation of target priority and sensor- target pairing, the index function is established. On the basis of linear programming, the distribution algorithm of sensor resources is given which is used for detection, tracking, and identification of target. ? Based on probability statistics model and information entropy, a method of sensor resource distribution used for detection and classification is put forward. ? Based on state dynamic model and information entropy, a method of sensor resource distribution used for detection, tracking, and classification is proposed. At the same time, an approach to dynamically determine priority of detection, tracking, and classification is presented. ? On the basis of discrimination, the index function of tracking system is computed, and sensor resource is distributed by using linear programming. Then, a joint multitarget probability is defined. Because the joint probability is constrained by measurement and Markov transition updates, the information gain fore-and-aft measurement can be computed in discrimination. Thereby, the efficient distribution of sensor resource is realized. 5. In consideration of the applicable environment of multisensor systems, the modeling method of information warfare is initially discussed, and a prototype of information warfare system is given.