Research On Direction Finding Method Of Colocated MIMO Radar

In recent years, a new-style radar system was presented, called as MIMO radar which was motivated by the attractive MIMO communication theory. MIMO radar has two types,both statistical MIMO radar and colocated MIMO radar. It has become a hot research topic owing to a number of potential advantaged over the conventional phased-array radar. This paper mainly research direction finding of colocated MIMO radar included both monostatic MIMO radar and bistatic MIMO radar, based on compressed sensing theory and spatial spectrum estimation theory.According to Shannon sampling theorem which is the basis of modern signal processing,signal can be reconstructed when the sampling frequency is higher than Nyquist frequency.However, based on compressed sensing theory, the signal can be still reconstructed perfectly if the sampling frequency is lower than Nyquist frequency. Since targets in airspace have sparse property which is the basis of compressed sensing theory, compressed sensing theory can be applied to direction finding. For coherent sources and uncorrelated sources, in bistatic MIMO radar, a CS-root MUSIC algorithm is presented in this paper. The receiving data is divided into signal and noise subspace using singular value decomposition. Then this idea keeps signal subspace to reduce both computational complexity and the sensitivity to noise,and signal subspace is reconstructed. Reconstructed signal subspace is only related to DOA of target, which can reduce the number of dimension of redundancy dictionary matrix. Finally,constructing redundancy dictionary matrix only related to DOAs and weighted matrix based on Capon spectrum to increase accuracy, DOA of target can be estimated by compressed sensing theory. According to orthogonality between steering vector of the target and noise subspace, DOD of the target can be estimated by root MUSIC algorithm. This method has better angle estimation performance, and the target's DOD and DOA can be paired automatically. In monostatic MIMO radar, for coherent sources and uncorrelated sources, a sparse representation scheme for DOA estimation is proposed, and the data only have a column. The theoretical basis of this idea is that eigenvector of big eigenvalue obtained by eigen-decomposition of the receiving data is linear combination of target's steering vector,and coefficients of this linear combination can be reconstructed by compressed sensing theory.If targets are coherent sources, the theorem still hold for eigenvector of maximal eigenvalue.Therefor DOA of target can be estimated by a sparse representation scheme using eigenvector of maximal eigenvalue. Since steering vector of the target in monostatic radar contain redundancy elements, a reduce-dimensional transformation matrix is utilized to eliminate redundancy elements of eigenvector of maximal eigenvalue. Then a weighted matrix based on orthogonality between steering vector and noise subspace is used to increase the estimation accuracy. In contrast with existing algorithm based on compressed sensing theory, this method only has one column, and has better accuracy.Since ESPRIT algorithm in bistatic MIMO radar have bad accuracy in low signal-to-noise ratio, a generalized ESPRIT algorithm is proposed in this paper. For incoherent signal sources, columns of array manifold and eigenvector of big eigenvalue span the same subspace. Signal subspace is utilized to reconstruct array manifold, and DOD of target is estimated by rotational invariance property of array manifold. Then corresponding array manifold is reconstructed by signal subspace which rows are exchanged by a transformation matrix, and DOA of target is estimated by the rotational invariance property of corresponding array manifold. Simulation results verify the effectiveness of this method. A generalized ESPRIT-root MUSIC algorithm is presented in this paper. While DOD of target is estimated by the generalized ESPRIT, DOA of target is estimated by root MUSIC algorithm.In contrast with ESPRIT-root MUSIC algorithm, this method has better performance in low snapshots and signal-to-noise ratio.