Research Of Maneuvering Target Tracking Based On Interacting Multiple Model And Smooth Variable Structure Filter

With the increasing complexity of modern target tracking scenarios,the measurement uncertainty brought by the unknown tracking environment and the modeling uncertainty brought by the high maneuvering target put forward higher requirements for maneuvering target tracking technology.This dissertation focuses on these two uncertain factors which affect the accuracy and stability of the maneuvering target tracking system.By using the idea of interacting multiple model and introducing the theory of smooth variable structure filter,the algorithm of single and multiple maneuvering target tracking is deeply studied.1.The smoothing variable structure filter proposed in recent years is introduced and its advantages over other traditional filtering estimation algorithms are analyzed.Although the standard smooth variable structure filter sacrifices some filtering accuracy,it shows good filtering robustness in high noise filtering environment with uncertain model.According to the standard smooth variable structure filter,the smooth variable structure filter that complements the state error covariance and the time-varying smooth boundary layer is further derived,which makes preparation for the subsequent construction of new tracking strategies.2.In order to eliminating the deficiency of tracking stability of traditional interactive multi-model algorithm,the IMM-SVSF algorithm combining smooth variable structure filter with interacting multiple model is introduced.Because SVSF itself is a suboptimal estimation algorithm,IMM-SVSF sacrifices some filtering accuracy while improving the filtering stability.When the probability of uncertainty in the tracking environment is low,the advantage of robustness of IMM-SVSF is not obvious.Therefore,on the basis of IMM-SVSF,this dissertation further proposes the IMM-EKF-SVSF algorithm.This strategy divides the strength of maneuverability of the preset model firstly,and makes EKF and SVSF participate in different ways according to the strength of maneuverability.IMM-EKF-SVSF inherits the high accuracy of EKF and the robustness of SVSF effectively,so that the algorithm itself has better adaptability and achieves excellent single maneuvering target tracking effect in the tracking scenario with measurement uncertainty.3.In order to further improve the tracking stability of multi-maneuvering targets,this dissertation studies the Gaussian mixture probability hypothesis density filter algorithm which is based on the theory of random finite set,then combines it with the IMM-EKF-SVSF proposed previously,and put forward the IMM-EKF-SVSF-GMPHD algorithm that can be applied in the multi-maneuvering target tracking scenario.The algorithm is based on the framework of Gaussian mixture filter,and the recursive process of many Gaussian components is processed by multi-model filtering in parallel.The filtering results of multiple models are weighted and summed to obtain the final filtering result.This algorithm utilizes the state estimation and number estimation of GMPHD for multi-target effectively,and also eliminates the shortcoming of GMPHD's inability to track maneuvering targets.It uses IMM-EKF-SVSF reasonably to track multiple maneuvering targets,and shows excellent tracking performance in multi-target tracking scenarios with model uncertainty.