Research On Sensor Control Strategy Based On Probabilistic Hypothesis Density Filter

With the rapid development of science and technology,a large number of high-performance sensors and a variety of advanced information processing technology have been put into use,which makes it possible to realize the deep level sensor control.Meanwhile,due to the urgent demand of modern warfare,and increasing complexity of multi-source information fusion system,various target tracking system are put forward higher requirements on sensor control,and the theories and methods related to sensor control are facing more severe challenges.Therefore,the study of sensor control strategy has very great theoretical value and reality-oriented meaning.The multi-target tracking method based on random finite set(RFS)has been widely concerned,because it can avoid the data association in traditional multi-target methods.Probability hypothesis density(PHD)filter is one of the most representative approximation multi-target filter.Based on the theoretical system of RFS,the dissertation mainly investigates the multi-target tracking sensor control method,and the main research contents are as follows:1)In allusion to the sensor control problem in multi-target tracking,under the framework of Partially Observable Markov Decision Processes(POMDPs),firstly,the general method of sensor control based on information theory is given through random finite set modeling.Secondly,the expression of Cauchy-Schwarz(CS)divergence is deduced based on the statistical assumptions of the probability hypothesis density filter and the probability of the cardinalized PHD(CPHD)filter.Using CS divergence as the sensor control evaluation function,and then calculating the information gain in multi-target filter recursion process.2)Combined with multi-target tracking problem,the implementation of Sequential Monte Carlo Probability Hypothesis Density(SMC-PHD)filter is given,and the first-order statistical moments of multi-target are expressed by weighted particles,at the same time,heuristic method is used to select the set of sensor control scheme,based on the above-mentioned multi-target CS divergence represented by multi-target statistical moment,the specific values of information gain of different decision-making scheme in sensor control set are solved.Then,selecting the final sensor control scheme by maximizing the information gain as the evaluation criteria.The simulation results show that the sensor control algorithm based CS divergence can adjust the sensor itself according to the current filter results,and then completing the goal of making the global target tracking performance optimal.The effect of multi-target tracking control scheme is superior to several other control schemes.3)A sensor control method based target threat degree is presented.In this paper,the factors that affect the threat degree of the battlefield environment are analyzed,in order to quantify the influence of target distance,heading and velocity on the target threat degree,the tactical significance map(TSM)function is established.Meanwhile,determining the target which has the maximum threat degree at different times by using TSM function values.And the CS divergence of the maximum threat degree target in the recursive filtering process is taken as the evaluation function,then,achieving sensor control based the target threat degree by maximizing the information gain as a criterion.The simulation results verify the effectiveness of the method.