Research On Multi-radar Coalition Game Theory And Methods For Intelligent Cooperative Target Tracking

In order to adapt to the increasingly complex military and civilian multi-target application scenarios,the random finite set(RFS)method for multi-target tracking has been vigorously developed in recent years.Based on the finite set statistical(FISST)theory,the original RFS method proposed a systematic Bayesian multi-target tracking framework that is capable of estimating the unknown and time-varying number and state of multitargets from complex environments such as clutter.Unfortunately,the tracking trajectories formed from target state estimation suffer from the problem of trajectory fragmentation/discontinuity.Thus,in order to be able to directly estimate continuous and smooth target trajectories in principle,the RFS method has been recently developed as the trajectory RFS theoretical method,which is capable of accomplishing direct filtering from sensor measurements to multi-target trajectory estimation.This dissertation takes the trajectory RFS method theory as the research background,and uses the most computationally efficient moment approximation filters in the trajectory RFS method,i.e.,the trajectory probability hypothesis density(TPHD)and the trajectory cardinality probability hypothesis density(TCPHD)filters,as the research topic.The algorithmic research on the moment approximation trajectory set TPHD and TCPHD filters for the application problems: spawning targets,maneuvering targets,target joint tracking and classification,was focused on.The main innovations are as follows:1.For the problem of spawning multi-target tracking,three trajectory spawning models,i.e.,the Bernoulli,Poisson,and zero-inflated Poisson models,are constructed.The trajectory set dynamic model accommodating the spawning motion of the target is proposed,and moment approximation trajectory set TPHD and TCPHD filter predictions corresponding to the different trajectory spawning models are derived on the basis of trajectory set spawning dynamic modeling.Then the linear Gaussian mixture(LGM)implementation of this filter prediction is developed.Combining the standard moment approximation TPHD and TCPHD filter updates,the proposed algorithm has the ability to spawning multi-target trajectory tracking,which is referred to as the S-TPHD and STCPHD filters.Simulation experiments are conducted to verify the effectiveness of the proposed algorithms by designing a multiple spawning target motion scenario,and the performance differences between the different spawning models are discussed.2.For the problem of maneuvering multi-target tracking,the trajectory motion model variable(TMM)and the augmented trajectory set are proposed to describe the trajectory maneuvering characteristics.The system tracking model of this augmented trajectory set is constructed.And the multiple model(MM)extensions and TMM extensions of the moment approximation trajectory set TPHD and TCPHD filters are successively derived on the basis of the augmented trajectory tracking modeling,which are referred to as the MM-TPHD,MM-TCPHD and the TMM-TPHD,TMM-TCPHD filters.The proposed algorithms are capable of maneuvering target trajectory tracking and motion model estimation,and since the MM extension is a motion model marginal approximation of the TMM extension,the LGM implementation of the TMM extension is subsequently mainly developed.Simulation experiments analytically verify the tracking performance of the proposed algorithms by designing a multiple maneuvering target motion scenario.3.For the the problem of multi-target joint tracking and classification recognition,the category trajectory variable and category trajectory sets based on motion model classification strategy are proposed to describe the trajectory state and category information of the target.The trajectory set dynamic motion model and the measurement model are elaborated in detail by constructing the tracking system model of the category trajectory set.Then the moment approximation trajectory set TPHD and TCPHD filters with ability of multi-target joint tracking and classification(JTC)are derived on the basis of the category trajectory tracking modeling,referred to as JTC-TPHD and JTC-TCPHD filters,whose LGM implementations are subsequently developed.Simulation experiments analyze and verify the multi-target JTC performance of the proposed algorithms by designing a multi-target motion scenario in which some unmanned aerial vehicles(UAVs)and aircrafts coexist.