| Cognitive radar has obvious advantages over traditional radar.It can dynamically adjust working mode,signal waveform and data processing algorithm according to the change of target environment.It is one of the important directions of next generation intelligent radar research.This dissertation mainly studies the target tracking theory of cognitive radar and the key technologies involved.1.In the theory of cognitive radar tracking,(1)the working mechanism of cognitive functions in human brain,such as perception-action cycle,memory,attention and intelligence,and the theoretical inspiration of these functions for radar target tracking are studied.Perception-action cycle is the basic mechanism to construct the closed-loop tracking loop of cognitive radar environment perception,information feedback and waveform transmission.The functions of memory updating and activation are helpful to inspire radar to learn environment,store information and call database or auxiliary knowledge.Based on the selective filtering and feature integration mechanism of attention,radar can make more efficient use of limited resources for fast target processing.Intelligence requires that radar has strong learning and adaptive ability,and can make timely and accurate decisions or actions.(2)Aiming at the three basic problems of target motion state perception,cognitive waveform transmission and cognitive reception processing,firstly,the mathematical expression of the error ellipsoid volume used to describe the perceptual uncertainty of the target is derived from the geometric structure and physical meaning of error covariance estimated by Kalman filter algorithm.Then,based on the idea of perception-action cycle,the target tracking method of waveform selection based on minimum information entropy criterion is proposed.Finally,the problems caused by the uncertainties of target motion state,measurement origin,measurement equation nonlinearity in radar data processing are studied,and the cognitive solutions are given.2.In the key technology of cognitive radar tracking,maneuvering target tracking,data association,particle filter and joint detection and tracking are studied in detail.Waveform selection technology based on minimum information entropy criterion runs through it,and realizes the joint cognition of radar receiver and transmitter.(1)For maneuvering target tracking,firstly,the Cramer-Rao Lower Bound(CRLB)expression of joint range-range rate and azimuth measurement error covariance based on transmitting waveform and antenna structure is given.Based on the current statistic(CS)model,the joint waveform and angle optimization algorithm for maneuvering target tracking is studied.Then,for the design defects of traditional CS model,constant velocity(CV)model is introduced to interact with CS in the framework of Interacting Multiple Model(IMM).Inspired by the three-stage information processing mechanism of human brain memory,a time-varying model transfer probability IMM(TIMM)algorithm with embedded memory is proposed,which overcomes the unnecessary competition between models caused by fixed model transfer probability.Finally,in the process of waveform selection using TIMM,the integrated prediction error covariance obtained by weighting the prediction position with predicted probability is used for information feedback,which avoids the situation that the waveform is easily trapped into local optimum by using only the feedback information of a single model.(2)For data association,firstly,based on the visual selective attention mechanism,a comprehensive TIMM(CTIMM)adaptive association gate design method is proposed.The center and size of the gate are obtained by weighting predicted position with predicted probability of CS and CV models,so that the association gate can be adaptively adjusted according to the target maneuver,and the radar computing resource consumption and target tracking performance are well considered.Then,based on the mechanism of visual attention feature integration,an optimized probabilistic data association(OPDA)algorithm is proposed.This algorithm classifies the common measurements in the cross-section of association gates using the integration of target location and motion features,and transforms the multi-target data assocation problem into multi-single-target data assocation problem,which enhances the environmental adaptability of traditional probabilistic data association(PDA)algorithm.Finally,the modified Riccati equation is used to estimate the filtering error covariance of each waveform,and the waveform at the next moment is adaptively selected according to the minimum information entropy criterion to improve the tracking performance.(3)For particle filter,firstly,in view of the problem that traditional particle filter often neglects the uncertainty in target tracking and presupposes a fixed number of particle samples in advance,which may result in low accuracy due to too few samples or low efficiency due to too many samples,the lower bound expression of particle sample number under the condition of expected filtering accuracy is derived based on the uncertainty error ellipsoid theory.An adaptive particle filter(APF)algorithm based on information entropy is proposed.Then,in order to reduce the complexity brought by the combination of waveform selection and APF algorithm,a parallel cognitive structure particle filter algorithm based on APF and Extended Kalman Filter(EKF)is proposed.In this algorithm,APF provides EKF with the initial value of filtering every time to avoid divergence of EKF algorithm.EKF provides the best matching measurement for APF algorithm through waveform selection,so that APF can always obtain the best tracking performance with the minimum number of particle samples.(4)In the aspect of joint detection and tracking,firstly,the theory of time delay-Doppler resolution cell is deeply analyzed.The approximate expressions of average detection probability and measurement error covariance of Linear Frequency Modulation(LFM)pulse are given.Then,aiming at the non-linear moving target measured by range-range rate and azimuth,a time delay-Doppler and azimuth joint resolution cell with "prism" structure is designed,and an APF/EKF parallel structure detection threshold and waveform adaptive tracking algorithm are proposed,which takes both tracking efficiency and tracking quality into account. |
PDF Full Text Request