Research On Random Matrix Methods For Robust Target DOA And DOD Estimation In MIMO Radar With Large Arrays

As a new kind of radar,multiple-input multiple-output(MIMO)radar has attracted wide attention from scholars both at home and abroad,since it has many advantages compared with the traditional radar in target detection,identification and localization.In bistatic MIMO radar,the target joint direction-of-arrival(DOA)and direction-of-departure(DOD)estimation is the prerequisite for the target location.In recent years,its signal processing methods have been widely studied.The previous methods for joint target DOA and DOD estimation in MIMO radar have relied on the hypothesis that the number of snapshots grows large comparatively to the number of arrays,which would lead to the degradation in performance when the number of snapshots is not enough.With the growing maturity of MIMO radar technology,the large array system has become a development trend in the future.In this case,the number of arrays is comparable with the number of snapshots,or even larger than the number of snapshots.In this situation,the method of replacing the statistical mean with the time average to estimate the sampling covariance matrix is no longer hold.In addition,the noise in actual environment may show the non-Gaussian characteristics and present heavy-tailed distribution.In this condition,the accuracy of the traditional methods can be greatly reduced and the DOAs and DODs can even not be estimated.In this thesis,the random matrix theory,beam space theory and linear shrinkage technology are used as tools to further investigate robust target DOA and DOD estimation methods for MIMO radar with large arrays.The work is supported by the National Natural Science Foundation of China about"Robust target detection and estimation for MIMO radar based on large dimensional random matrix theory”(No.61371158).The innovative work of this thesis is as follows:According to the condition that the number of arrays and the number of snapshotsare the same order of magnitude,the signal model in element space for bistatic MIMO radar is established in this thesis.A new method called 2D-spikeMUSIC is proposed based on random matrix theory.On the basis of the spiked model,the consistent estimator of the covariance matrix is generated by the Stieltjes transform of Marcenko-Pastur distribution.Further,the robust target DOA and DOD estimation is realized.Simulations show this method is suitable for the large array system which can effectively estimate the target DOA and DOD under the condition of large arrays.In view of large amount of computation in random matrix method,the signal model in beam space for MIMO radar is established in this thesis.A new method called 2D-beam-GMUSIC is proposed which combines beam space and random matrix theory.The received signals are converted from the element space to the beam space by DFT weights,and then the DOA and DOD are estimated using GMUSIC method.The computational complexity is surely reduced by reducing the dimension under the premise of ensuring the accuracy of the target DOA and DOD estimation in MIMO radar with large arrays.Aiming at the problem that the noise distribution may be non-Gaussian distribution and the number of arrays can be greater than the number of snapshots,firstly a new random matrix method called 2D-RMW is proposed using tools such as the contour integral technique and the residue theorem for Gaussian noise background.Simulations verify the effectiveness of this method.On the basis of this method,a robust estimation method called 2D-RLSMW for non-Gaussian noise background is proposed through the fixed-point iteration combining with linear shrinkage and random matrix theory.Also,it can be a robust estimator when the number of arrays is larger than the number of snapshots.