Research On Algorithms For Single Observer Passive Location And Tracking Based On Cubature Kalman Filter

Single observer passive location system has become the hotspot of research in the field ofelectronic warfare because of its significant advantages in self-hiding, simple equipment, relativeindependence and far-distance detection, etc. Since measurements are nonlinear function of thetarget state, single observer passive location and tracking technology is a nonlinear filteringproblem in substance and the reaserch for filter algorithms of high location precision, fastconvergence rate and strong stability according to tracking situation is a process which needs tobe continuously explored. The spatial-frequency domain information is used and the reaserch isbased on a new Sigma point filtering algorithm——cubature Kalman filter（CKF）in thisdissertation and the main contents are summarized as follows:1. The theory of single observer passive location based on the spatial-frequency domaininformation is studied based on particle kinematics. The range error is analyzed. The variousfactors affect the range accuracy of the location method and how these factors influence thelocation accuracy are researched. Observability of the target is the precondition of trackingalgorithms, the pseudo-linear process for the measurement equation is used and observabilitycondition of constant velocity target and two classes of conventional maneuvering (constantacceleration and constant turn rate) targets are investigated for the fixed single observer passivelocation system based on the spatial-frequency domain information.2. The cubature Kalman filter（CKF） is applied to single observer passive location. Thecore idea of CKF is that the Spherical-Radial rule is directly used to calculate the mean andcovariance of the nonlinear random function and the implementation of the method is simple andhigher accuracy of state estimate is achieved. Simulation results show that CKF is a nonlinearfiltering algorithm with excellent performance in single observer passive location and tracking.Taking the low measurement precision of single observer passive location system intoconsideration, a backward-smoothing CKF（BSCKF）which combines backward-smoothing withthe cubature Kalman filter is proposed. In this algorithm, the backward-smoothing result is usedto the recursive cubature Kalman filtering to improve the performance of location. Simulationresults show that BSCKF is better than CKF in location precision and convergence speed.3. A strong tracking cubature Kalman filter（STCKF） is proposed according to “current”statistical model for the sudden maneuver case. This algorithm improves the adaptive trackingperformance by introducing a fading factor to filtering process through adjusting the errorcovariance of the predicted state, the error covariance of the predicted measurement and theassociate covariance of the predicted state and measurement to adjust the filter gain matrix on-line. Simulation results show that when there is only common maneuver the performance ofSTCKF and CKF are nearly the same, whereas there is a sudden maneuver，the performance ofSTCKF is much better than that of CKF. Considering the situation of maneuvering targettracking using IMM in single observer passive location system, a measurement updateCKF-IMM algorithm is proposed. This algorithm uses Markov process to control the switchingamong the sub-models, and CKF for filtering of each model. The outputs of all parallel CKF areweighted sum as an integrated estimation and the integrated estimation is put through thenonlinear measurement update for further approaching to the real value by reducing themeasurement error influence. Simulation results show that, CKF-IMM has lower estimation errorcomparing with the EKF-IMM and UKF-IMM and the measurement update CKF-IMM is ofhigher location precision than CKF-IMM.4. The precision and stability of the filter could be influenced owing to the fact that outliersmay occur in sequential measurements for single observer passive location system, and the filtermay be diverged seriously. For the sake of this reason, combining with the scaled—contaminatednormal distribution mode（lSCNM）, a robust cubature Kalman filter is proposed based on Bayestheory. The measurement error is set up by the SCNM based on CKF and the measurementprediction residual error variance matrix is adjusted adaptively according to the posteriorprobability of outliers. Simulation results show that, this algorithm with high location precisionand strong robustness is effective at the aspect of reducing the impact of discrete or continuousoutliers in measurements.