Robust eigenstructure-based DF in the presence of mutual coupling

For over a decade, considerable effort has been focussed on the array-based direction finding (DF) problem by the signal processing community. The primary concern has been the development of reliable superresolution algorithms for resolving two or more simultaneous signals incident upon the array at a separation smaller than the Rayleigh resolution of the array. These algorithms (such as MUSIC and WSF) rely inherently on a complete knowledge of the array's response to "ground truth" signals, arriving one at a time, from all directions. This array manifold represents the behavior of the array in its surroundings. Generating the array manifold is commonly referred to as array calibration. This time consuming task is not suited for mass produced arrays or systems in need of constant recalibration. In fact, for some tactical military DF systems, setting up known calibration sources is often impossible.; In practice, it is well understood that DF systems do not perform nearly as well as theory predicts. This is often due to sensor gain/phase errors and mutual coupling between the elements of the array. Many of these issues have been relegated to the domain of "systems engineering" and until recently were not addressed in a systematic, theoretical manner. To address these significant problems that prevent achieving the promise of theoretical DF performance in practical systems, this dissertation addresses and answers two primary questions: (1) In the presence of mutual coupling, can the performance problems of standard DF algorithms be remedied via reduced sensitivity to errors in array calibration? (2) Can a reliable method of array calibration be devised, which (a) can be performed while the DF system is in use, and (b) does not require known calibration source locations? Simulations and analytical results show that the answer to both questions is yes. The ability to control the array impedance via the impedance of the element terminations (say, by a factor of 10) leads to performance improvements that can achieve a 10-100 fold increase in the mean-squared error (MSE) estimates of signal DOAs, using either MUSIC or WSF. The ability to control the array impedance allows initial estimates of signal DOAs, in the presence of unknown mutual coupling. These estimates are then used to bootstrap an autocalibration algorithm--that corrects and extends the earlier autocalibration method of Pierre and Kaveh (Pie91) --which achieves a 100-fold improvement in array calibration residual MSE.; This work is an important step in more fully categorizing the effects of mutual coupling upon antenna array calibration. This has potential of producing DF systems which have a robust capability in the presence of side effects, such as array degradation and the dynamic array environment. By introducing an improved signal model for the DF problem, developing a technique for reduced mutual coupling, and solving the unknown sources-of-opportunity calibration problem, this thesis has produced a framework for robust eigenstructure-based DF.