Study On Angle Estimation Algorithm For MIMO Radar

MIMO radar has been widely investigated in recent years, owing to its advantages in target detection and parameter estimation. By transmitting orthogonal waveforms at the transmit site and performing matched filtering for all orthogonal waveforms at the receive site, MIMO radar can obtain the transmit degrees of freedom (DOF) and increase the system DOF, which is beneficial to improve angle estimation performance. As the size of the matched filtered data is proportional to the product of the number of transmit sensors and the number of receive sensors, the computational burden on parameter estimation and the number of the required training data are very large, especially in two-dimensional (2-D) angle estimation. Therefore, the research on the fast angle estimation algorithms with lower requirement on temporal snapshots is of importance in practice for MIMO radar.This dissertation mainly focuses on the angle estimation algorithms for MIMO radar. The research of this dissertation is summarized as follows:1. The signal model of MIMO radar is first established. According to the principle and the signal processing procedure of MIMO radar, a sufficient statistic is given. Then some concepts including the virtual sensor, the DOF and the maximum number of resolvable targets and several virtual arrays of typical array configurations are analyzed. Finally the Cramer-Rao bound (CRB) of the two-dimensional angle estimation for MIMO radar is derived and the simulation results are given.2. Angle estimation for monostatic MIMO radar is researched. The conventional angle estimation technique, such as Capon spatial spectrum method, is applied to MIMO radar, and one-dimensional (1-D) and 2-D angle estimation of MIMO-Capon spatial spectrum for monostatic MIMO radar are presented. Meanwhile, the maximum number of resolvable targets for uniform linear arrays (ULA) and L-shape arrays are verified by simulation. To avoid angle search in conventional MIMO-Capon spatial spectrum, an angle estimation algorithm for monostatic MIMO radar using polynomial rooting is proposed, which alleviates the temporal snapshots requirement and the computational burden resulting from covariance matrix estimation and inversion of the covariance matrix. Utilizing the redundancy of the received data, reduced-dimensional processing is firstly performed. To further alleviate the computational burden, a polynomial rooting method is proposed to identify the extreme points of Capon spatial spectrum. The presented method implements angle estimation in a reduced-dimensional space, and thus owns lower requirement for temporal snapshots and computational burden compared to the full-dimensional processing.3. Angle estimation for bistatic MIMO radar is researched. A direction finding algorithm based on Rayleigh quotient is first presented. In the presented method, the property of the kronecker product is first utilized to reformulate the MUSIC spatial spectrum in the quadratic form. Then, by constructing the orthogonality constraint of the target angles to the quadratic form, all target angles can be estimated by solving the constrained quadratic form with a 1-D search. Moreover, the 2-D directions can be automatically paired. The simulations verify that the proposed method has good angle estimation performance at low SNR when the number of snapshots is small. However, the method based on Rayleigh quotient which needs angle search in full angle region still has heavy computational burden when the probable locations of targets are unknown. Therefore a method for multi-target localization based on polynomial rooting is developed in bistatic MIMO radar. Utilizing the property of transmit-receive steering vector, the 2-D direction finding is transformed into two 1-D direction finding procedures. Then the polynomial rooting technique is employed to determine the 1-D direction. The proposed method avoids the conventional 2-D spectrum peak searching, and the estimated parameters are automatically paired without additional paring computation. Numerical results verify the effectiveness of this method. Finally, performance comparisons between the two proposed algorithms are made in simulation. The method based on Rayleigh quotient has good angle estimation performance at low SNR when the number of snapshots is small, while the method based on polynomial rooting owns low computational burden.4. Low elevation estimation for MIMO radar is researched. A fast algorithm for low elevation estimation in multipath environments for MIMO radar via matrix pencil is proposed. The signals both in transmit multipath and receive multipath are considered. A forward-backward matrix pencil is formed based on the single-sampled vector to alleviate the effect of noise in low signal-to-noise case. After reduced-dimensional processing for the matrix pencil, the generalized eigen-decomposition is employed to estimate the low elevation in multipath environments directly. The proposed method can overcome multipath effect effectively in case of low SNR and single secondary data and estimate multi-target simultaneously. The proposed method avoids spectrum peak searching and reduces the computational burden. Numerical results verify the effectiveness of this method. 5. Low elevation estimation in the presence of multipath effect and mutual coupling error for MIMO radar is researched. A method for low elevation estimation in the presence of multipath effect and mutual coupling error for MIMO radar is proposed. Mutual coupling and multipath effect are taken into account for the signal model. Utilizing the characteristic of the mutual coupling matrix of uniform linear arrays, derivation of the subspace-based self-calibration algorithm is given. The algorithm can obtain the estimated angle as well as the reflectivity and the mutual coupling matrix simulataneously. Additionally, the CRB is derived. Simulation results illustrating the performance of the algorithm and comparison with the CRB are also presented.