Study On Subarray Clustering And DOA Estimation

Monopulse estimation and direction of arrival (DOA) estimation based on subspace are two techniques for one-source location and multiple-source location respectively. And the corresponding key problems in their applications are to generate sum and difference patterns with required features, such as low sidelobe level (SLL), high directives, and to improve robustness of the algorithms to kinds of non-ideal situations respectively. For monopulse estimation at subarray level, joint optimization of subarray configuration and weights is discussed in this dissertation according to the given sum and difference patterns or their features; Robust DOA estimations in the presence of strong jamming or signals, array errors and unknown source numbers are studied for subspace-based estimation. The main contributions and innovations are summarized as following:1. For mitigate the amplitude of the grating lobes in difference pattern due to the subarray beamforming, that both the sum and difference beam patterns are generated at subarray level, is proposed instead of the available model that the element weighting produces the sum pattern and subarray weighting give the difference pattern, which is just its special case that the all subarray weightings for sum channel are one. Then, the element weighting, sum subarray weighting and difference subarray weighting can determined by the desired sum and difference pattern together and a flexible compromise among sum and difference patterns can be reached.2. For the optimal sum and difference patterns synthesis at subarray level, a novel approach based on grouping to optimize subarray configurations and weights is presented. In the frame work of minimizing the distance between optimal and synthesized sum and difference excitations, that to which subarray an element belongs is adjusted according to a distance defined from an element to a subarray and iterative subarray partition is reached. It overcomes the shortcoming that available iterative methods can not be applied to planar antenna arrays. Meanwhile, by exploiting a logarithm-transformation provided low SLLs sum and difference patterns based on tapering, the original non-convex optimization problem with a given subarray configuration can be simplified to a convex optimization, and element weighting and the two weighting for sum and difference beam at subarray level, can be obtained together without iteration needed. And a control on the maximum attenuation taken by element weighting can be added to the former optimization without changing its convex nature, and its quantized attenuation values can be realized by an iterative optimization.3. For the same problem above, a new method in the frame work of pattern synthesis is presented, which using the SLLs controls as its constrains, the weighted sum of the mainlobe distortions of sum and difference patterns as its cost function, is solved by optimizing the element weighting vector and subarray weightings alternately. It can control the ratio shape of the difference and sum pattern function, including the slope, the linear extent of the magnitude of the sum and difference ratio, its phase response, which overwhelms that just the difference pattern slope in the null direction can be controlled in approaches available.4. For DOA estimation of weak signal in the presence of strong jamming, two new methods are given, based on extended noise subspace and power normalized respectively. The first approach eliminates the peaks of the strong jamming, by adding their steering vectors into the noise subspace, which reduces the computation burden and just needs the strong jamming number. The second, according to that stronger signal has greater attenuation while the weaker one has the less, transforms the antenna outputs through the eigenvectors matrix of correlation matrix and then normalizes each of the channel outputs according to their power themselves and does not require any pre-knowledge of strong jamming.5. For DOA estimation with array error, a joint estimation of DOAs and mutual coupling matrix based on ESPRIT algorithm is provided. It exploits the poverty of the coupling matrix to a uniform linear array (ULA), subtracts two subarrays with the same structures and coupling characteristic, estimates the DOAs using ESPRIT and then obtains the coupling parameters in the Least-Square (LS) sense. In the method no need to iterate, no problem not to converge and small computation burden are reached. And the computer simulation also shows that it's robust to sensor gain and phase uncertainties to a certain extent.6. For DOA estimation without the source number known, a joint estimation of DOAs and source number based on MUSIC is considered. Based on the fact that some unexpected peaks appear in MUSIC spectrum when the source number is over estimated, exploiting that steering vector of the real signal is orthogonal to the corresponding noise subspace, a function is constructed to pick the unexpected peaks out, then the estimation of real source number and DOAs is obtained. It can be looked as either a DOA estimation method without the source number known or a source number estimation method recurring to the DOA estimation, the computer simulation show that its source estimation is valid to the space color noise that is independent but with unequal power among different sensors.