The main objective of this dessertation is to develop an adaptive beamformer which is robust to array imperfections and is highly fast and numerically stable. In the presence of undesired interference signals, desired signal reception using an adaptive array system has been the subject of considerable interest and investigation because an adaptive array system provides the solution to separate signals that have overlapping frequency content but originate from defferent spatial locations. An adaptive array is a system consisting of an array of sensor elements and a processor coupled with adaptive control unit so as to pass a desired signal with minimum distortion while rejecting interference singals. Almost all of conventional beamformers so far studied require an underlying assumption that array characteristics are ecactly known. Unfortuatedly, even if slight errors in array characteristics occur, the linearly constrained beamformer not only cancells the desired siganl, but also tends to increase the norm of the weight vector, resulting in a poor performance. Array imperfections are caued by the gain errors of the phase shifters, or the mismatches between the nominal and actual values of sensor positions or sensor gains. Array uncertainty is a very important issue in practical application of adaptive array systems. Robustness to array imperfections is a highly desirable quality for any algorithm to possess, especially for hardware implementation. In this dessertation work, we investigate the causes of imperfect array problems and suggest several schemes to overcome imperfect array problems. First, we find a fact that the performance degradation of the linearly constrained beamformer with imperfect arrays is due to the rational that under the mismatch conditions the assumed steering vector of the desired signal is no longer orthogonal to the noise subspace of the array covariance matrix. This cause of imperfect array problems leads to develop a robust beamformer...