Research On Sensor Management Method Based On Generalized Label Multi-Bernoulli Filter

Sensor management technology has always been a hot research topic in the field of multi-target tracking.With the rapid development of multi-source information fusion technology,sensor management technology faces enormous demands and challenges.Therefore,the research on sensor management technology has great theoretical significance and engineering value.Sensor management problem in multi-target tracking,which is essentially a process of time-varying joint decision making and estimation.In the research,the optimal tracking control process is generally used to control the limited sensors to obtain the target tracking information to the greatest extent.In recent years,the sensor management method based on Finite Set Statistics(FISST)has attracted much attention due to its solid mathematical foundation and deep research background.In view of this,based on the finite set statistical theory,this paper uses the generalized label Dobernoulli filter to conduct in-depth research on sensor management in multi-target tracking.The main research contents are as follows:1)Aiming at the sensor management problem in multi-target tracking,based on the FISST theory,the corresponding sensor management methods are studied under the framework of Partially Observed Markov Decision Process(POMDP).First,a BOX-Generalized Labled Multi-Bernoulli(BOX-GLMB)filter based on the implementation of the box particle implements a multi-target state filter estimate.When the sensor management scheme is solved,according to the Rényi information increment and the multi-objective Bayesian filter expression,Based on Markov characteristics,the evaluation function of Bayesian filter based on Rényi information increment is derived.Then the interval sampling method is designed to sample the multi-objective probability density,and a set of weighted particles is used to approximate the multi-objective statistical properties,and the particle weights are updated by the box particle resampling,and then passed the Monte Carlo(MC)method.The expression of the BOX-GLMB filter based on the Rényi information increment is approximated.Finally,the solution steps of the sensor management evaluation function are given and simulated.The effectiveness of the proposed strategy is verified by comparing the tracking performance of differentsensor management strategies,target number estimation,program running time and sensor control trajectory cloud.Simulation experiments verify the effectiveness of the algorithm.2)A battlefield single sensor management algorithm based on Rényi information increment and Tactical Significance Map(TSM)function is studied.First,the target tactical importance rendering function(TSM)is used to measure the threat level of the target.Secondly,a multi-target tracking filter is used to estimate the target state using a generalized label Dobernoulli filter(BOX-GLMB)of the box particle implementation.Then,the TSM function value of each target is calculated according to the filtering state,and the target with the largest TSM function value is selected as the target with a greater threat degree.Finally,the Rényi information increment is used as the sensor evaluation function to maximize the information gain of the largest threat target.The sensor management criterion is to implement single-sensor management based on the battlefield environment.Simulation experiments verify the effectiveness of the proposed algorithm.