Array Self-Calibration For MIMO Radar With Gain-Phase Error

In recent years, with the intensive research in MIMO radar, it is more obvious that MIMO radar has great advantages over traditional phased array in target detection, parameter estimation, interference cancellation and low probability of interception. As MIMO radar utilizes more active devices in the transmitter and receiver, of which the gain and phase may change significantly due to machining error, working environment and service time. However, most MIMO radar signal processing algorithms require the knowledge of array manifold. In the case where gain-phase error exists, the gain-phase error will alter the array manifold leading to great degradation in algorithm performances. Therefore research in array gain-phase error calibration is of both theoretical and practical value. According to the usage of active calibration sources, the gain-phase error calibration methods are divided into two categories, one is pilot calibration and the other is self-calibration. Pilot calibration requires the knowledge of calibration sources with known directions, thus this method is restricted in its usage. This thesis mainly focuses on gain-phase error self-calibration in MIMO radar. The main contents are as follows:1. Traditional MIMO radar gain-phase error iterative estimation methods optimize cost functions with respect to the gain-phase error parameters and source direction parameters. In order to achieve the extremum of the cost function, an iterative block coordinate descent algorithm is applied w.r.t. gain-phase error parameters and source direction parameters. However, these algorithms suffer from local optimum. In the meantime, when the gain-phase error appears to be large, the iteration may have long convergence time or even tend to diverge. On the other hand, traditional self-calibration method based on instrumental sensors needs exhaustive search over multi-dimensional parameter space. Even though it does not require iteration and has no convergence problems, it has high computational burden and poor accuracy. In order to overcome those problems, this thesis proposed an improved MIMO radar gain-phase error self-calibration method based on instrumental sensors. In this method, three fully calibrated instrumental sensors are utilized to form rotational invariant factors. By computing and decomposing the auto- and cross-correlation matrix of received signals, we obtain the target directions as well as the array manifold. Thus the gain-phase error can then be extracted from the manifold. Simulation results verify the high accuracy of the proposed method.2. The aforementioned gain-phase error self-calibration methods for MIMO radar all require joint optimization w.r.t. target parameters and gain-phase error parameters. However, in the case of airborne MIMO radar, the clutter power is significantly greater than the target. Therefore with the unknown gain-phase error, it is impossible to jointly optimize w.r.t. target parameters and gain-phase error parameters. Although the compressive sensing based gain-phase error self-calibration method utilize the clutter returns instead of target returns, it is likely to diverge with large gain-phase error presented. In order to solve the aforementioned problems, a MIMO radar gain-phase self-calibration method based on the geometry of clutter subspace is proposed in this thesis. By modeling the clutter subspace using prolate spheroidal wave functions(PSWFs) and incorporating the Maximum-likelihood approach, we obtain a closed form gain-phase error estimator. This method utilize the geometry of clutter subspace, therefore it does not need to solve the target parameters. In addition, it does not require the independent identical distribution property of clutter scatters across different range bins, which adds to its practical value. Simulation and analysis show its low requirements for range bins and superiority over the compressive sensing based method.