Joint Estimation And Identification Based On Expectation Maximization

Due to the rapid development of target maneuverability and electronic warfare technology,the unknown and time-varying parameters become the key limiting factor for the high-precision target tracking.In essence,such problem belongs to the coupling problem of estimation and identification: the state estimation error triggers identification risk while identification risk causes state estimation error due to modeling mismatch.In this thesis,the joint optimization algorithms of estimation and identification for target tracking and track segment association are proposed in the framework of the EM criterion,and the main contributions are as follows:1.Reliable identification of mass-to-drag ratio and accurate estimation of system state are important and coupled issues.The joint optimization algorithm based on EM iterative framework is proposed for tracking a reentry target with unknown mass-to-drag ratio.Here the state vector about target movement is regarded as the missing data and the unknown mass-to-drag ratio is treated as the parameter to be identified.In the E-step,the random particle sampling stratergy is utilized to approximate the likelihood function to deal with the inherited nonlinearity.In the M-step,the numerical optimization algorithm is applied to update mass-to-drag ratio.In the simulation comparison with the traditional augmentation algorithm,our proposed algorithm shows the improvement in both state estimate and parameter identification.2.Motivated by tracking a maneuvering target in the presence of electronic deception jamming,the estimation problem of stochastic system with unknown inputs in both the plant and sensors is presented.The joint optimization scheme of estimation and identification based on the expectation maximization criterion is derived.The system state and parameters of maneuver and deception are regarded as hidden data and parameter to be identified,respectively.In the E-step,the state estimation is obtained by the URTS smoother,which has the lower computation burden than particle smoother.Meanwhile the identification of unknown parameters is derived analytically in the M-step.Finally,an example of tracking a maneuvering target accompanied range gate pull-off is utilized to verify the proposed scheme.3.There are track breakages frequently due to target maneuver.Therefore,the track segment association(TSA)technique for judging the identity of two fractured segments is necessary for reliable information report.Actually,the model parameter and stitching start and end time instants are critical for TSA and coupled with each other in the situation of target maneuver.In this paper,considering a common situation of coordinated turn maneuvering,a novel scheme for TSA based on the ECM is presented.Firstly,the association cost of each track pair is calculated based on ECM.In the E-step,the state estimate is obtained,and in the CM-step,parameters to be identified are updated simply.Then,the 2D assignment algorithm is utilized to find the global optimal pairs.The proposed algorithm is shown effective via both simulation data and real data of ATC radar system.