Research Of Maneuvering Target Tracking Algorithm Based On Smooth Variable Structure Filter

Target tracking technique has been widely used in military and civilian fields since its birth,and has achieved fruitful results.In the field of modern target tracking,target motion characteristics are very different from the past,mainly in high maneuverability and complexity of motion models,which challenges the robustness of target tracking.1.From the perspective of improving robustness of tracking algorithm,the robustness of smooth variable structure filter is added to traditional tracking algorithms to construct new combination algorithms,which obtain the accuracy of traditional tracking algorithms and the robustness of smooth variable structure filter at the same time.2.The current statistical model is a more practical model for maneuvering target tracking,however,when a target maneuvers in the acceleration with a larger change range or in the form of sudden change of state,the tracking error of current statistical model will increase.In this thesis,the combination algorithm of classical Kalman filtering and smooth variable structure filtering is applied in current statistical model,which realizes a new algorithm,namely CS-KF-SVSF.The new algorithm solves the problem that the filtering error of current statistical model will increase when the strong maneuvering target is tracked.3.For improving the filtering robustness of tracking system with non-Gaussian noises,a new algorithm of combining smooth variable structure filter and Gaussian sum filter is proposed,but the indicative effect of time-varying smoothing boundary layer decreases,which leads to the reduction of filtering accuracy.Based on this problem,a new strategy with a parallel structure is proposed,namely the modified GS-KF-SVSF.In the new strategy,the time-varying smooth boundary layer of smooth variable structure filter is used to indicate the conversion of filtering gain.And the new strategy achieves accurate and stable filtering results.The time-varying smooth boundary layer based on smooth variable structure filter has the function of indicating the uncertainties of modeling and the change of system parameters,which provides a new way for fault detection and enhancing robustness of system.