STAP Algorithm For Colocated MIMO Radar Based On Nystr?m Method

MIMO(Multiple-input multiple-output)radar is an attractive topic in the field of radar for its advantages,such as space diversity and waveform diversity.MIMO radar uses multiple antennas to send different(orthogonal or noncoherent)waveforms,and matched filterbanks in every receiver are used to distinguish different waveform components.Colocated MIMO radar is one that the distances between array antennas are close enough so that the target RCS and angel of arrive are equal for all the antennas to form a very long virtual array.The spatial and waveform diversity utilized in colocated MIMO radar can significantly increase the system degrees of freedom,then spatial resolution in separating the interference and target can be improved for Radar.Space Time Adaptive Processing(STAP)algorithm can suppress the strong jammers and clutters in receiving data of Radar while preserving desired target signal,so it can maximize the output SINR and improve the probability of detection for moving target.The STAP filter is a two-dimensional FIR filter,which jointly uses array spatial beamformer and temporal Doppler,so it can separate moving target from clutters in angle-Doppler domain.However,conventional STAP algorithm have large computation complexity and slow convergence,these two problems are limitaions for practice application.Nystr?m method can fastly approximate a high dimensional positive semidefinite matrix with its low-rank Nystr?m extension matrix.So in this paper,novel Nystr?m-STAP,AdaptiveNystr?m-STAP and their reduced-rank STAP approaches for colocated MIMO radar are proposed.In proposed algorithms,based on low rank property of clutter covariance matrix,clutter covariance matrix can be well approximated by Nystr?m-based covariance estimate(NCE)matrix with small number of clutter homogeneous training data support,then the eigenvectors of clutter subspace can be estimated to form STAP filter.In Nystr?m-STAP algorithm,choosing appropriate distinct rows of training data matrix to construct new data submatrix,the NCE is formed after the original training data matrix projected onto the space spanned by chosen rows.While for Adaptive-Nystr?m-STAP algorithm,adaptive column sampling technique is used to select informative columns of clutter covariance matrix,then NCE is obtained after scaling the selected columns and constructing intersection matrix,thus overcoming limitation of parameter selecting for Nystr?m-STAP algorithm.When the NCE is obtained in these two algorithms,the corresponding covariance submatrix of clutter-and-noise covariance estimate matrix can also be obtained by Nystr?m extension strategy.By the eigenvalue decomposition of low-dimension covariance sub-matrix,the clutter subspace can be estimated with high accuracy to improve output SINR and detection performance.When there are only clutter and noise interference components in cell under test,the STAP filter weight can be easily obtained from Eigen-canceler method.While for clutter plus jamming situation,combined with the unity decomposition of NCE and the block-diagonal feature of jammer covariance matrix,the STAP filter coefficients can be obtained by the inversions of small dimension matrix according to matrix inversion lemma,which can reduce the computational complexity.When the number of pulses in one coherent processing interval is large,in order to further improve convergence performance,the RR-Nystr?m-STAP and RR-Adaptive-Nystr?m-STAP are developed with the combination of Nystr?m method and Adaptive Displaced Phase Center Array technique.These algorithms choose successive and overlapped pulses in the training data,and the inversion of the low dimension interference covariance sub-matrix can be obtained with Nystr?m-STAP algorithm.After the processing of Doppler filter,the STAP filter can be formed.The simulation results illustrate our proposed methods can achieve better performance and lower computational burden.