Random Finite Set Method For Group Target Motion Modeling And Tracking

A group target consists of multiple sub-targets with a collaboration relationship.Multiple sub-targets that belong to the same group will present certain structural characteristics during the movement.When the resolution of the sensor is limited,the distance between the sub-targets in the group target is too close,which results in the group target showing indistinguishability,that is,the case where multiple sub-targets are located in a resolution unit.Group target tracking estimation mainly includes acquisition of target measurement,and filtering and estimation of the obtained measurement.In the research work of this paper,we describe the collaboration relationship among group targets with graph theory,and then apply the GLMB and Gibbs-GLMB filters to track and estimate the resolvable group targets and perform the structure and formation of the resolvable group targets.In-depth analysis and discussion,the structure of the resolvable group target and the method of determining the formation are given.Finally,the label and structure are combined to propose a structure label group target estimation algorithm.The main work includes:(1)State estimation of resolvable group state based on random finite set and graph theory.Firstly,the resolvable group targets are dynamically modeled by combining the graph theory,that is,the adjacency matrix represents the cooperative relationship between the group targets,and the CV and CT motion models of the group targets are established respectively.For group targets moving as CV model,GLMB filter and Gibbs-GLMB filter were used to track and estimate them.For CT motion,because CT motion is a non-linear motion,an Unscented Kalman filter(UKF)needs to be introduced.The UKF-GLMB filter and the UKF-Gibbs-GLMB filter are used to estimate.(2)Study on the structure and formation of resolvable group targets.In this section,after analyzing the structure and formation characteristics of group targets,we elaborated what is the structure and formation of group targets and their similarities and differences.Then,according to the characteristics of the two,methods are provided to determine whether the structure and the state of the group target have changed during the movement.(3)An analytical bayesian group target estimation algorithm based on RFS observation is proposed.Firstly,the single group target is defined in a discrete and countable space with a structure label.Then,the adjacency matrix is used to establish the cooperation model of the single group target,and the likelihood function of the single group target is given.