Research On Tracking Algorithms For Near Space Target With Strong Maneuvering

With the development of aerospace technology,near space aircrafts have drawn much attention from lots of countries due to their high speed,strong maneuverability and invisibility.It takes great challenges to the tracking system of radar with complex maneuverability of the near space targets.The study of the strong maneuvering targets tracking algorithms with high tracking accuracy and good real-time performance is becoming the focus of technical research about near space.This paper studies deeply the strong maneuvering target tracking algorithms of near space,and mainly focus on how to better match the motion characteristics of strong maneuvering targets and improve the real-time performance of filter algorithms.The main research work and main contributions of this paper are as follows:(1)The linear kalman filter algorithm and several nonlinear filter algorithms are studied in depth.The principle as well as advantages and disadvantages of various filter algorithms are summarized.Meanwhile,the performance of several nonlinear filter algorithms is compared and analyzed by the simulation,and the result shows that the unscented kalman filter algorithm is higher than other algorithms in the tracking precision and much less than others in the time complexity,which can adapt to the nonlinear maneuvering characteristic of the near space targets.(2)The modeling methods,along with the advantages and disadvantages of the basic motion models are summarized.On the basis of the idea of "current" statistical model,CS-Jerk model based on "current" statistics is established.It presumes the change rate of acceleration as a stochastic process with exponential autocorrelation and nonzero mean so as to make the estimation about the change rate of acceleration more accord with the actual motion process of strong maneuvering targets.(3)A modified Sine-Jerk model based on sine correlation function is proposed for the periodic and jumping maneuverability in cruise stage of near space targets.It models the change rate of acceleration as a stochastic process with sinusoidal autocorrelation,which improves the tracking accuracy of the periodic maneuvering targets in near space by the simulation.(4)In order to improve the real-time performance and tracking accuracy of the interacting multiple model algorithm,this paper applies the classification algorithm of logistic regression in machine learning to target tracking,and proposes the modified LR-IMM algorithm.The algorithm uses the logistic regression to pre-classify the motion state of the targets,and then selects a set of models that are,above a certain threshold according to the output value of classification,which establishes an new IMM algorithm with the adaptive model set.The simulation result shows that the new algorithm reduces the computational complexity of the standard algorithm,which improves the real-time performance.At the same time,the tracking accuracy is greatly improved due to the adaptability of the model set.