Reduced complexity adaptive structures for jam-resistant satellite communications

A linearly constrained wideband adaptive array sensor is described whose adaptive processor is realized with a two-dimensional orthogonal filter bank structure. It is demonstrated via simulation that the transform domain processor utilizing the computationally simple least mean square (LMS) algorithm is capable of convergence speeds which rival the more expensive to implement least squares methods. Furthermore, there is no additional cost in terms of the adaptive coefficient requirements (and only a modest increase in non-adaptive computational requirements) when compared to the standard tapped-delay-line (TDL) structure array utilizing the LMS algorithm. This performance increase provides a quicker approach to the optimal steady state solution, and is therefore more robust to a non-stationary environment than the corresponding TDL structure LMS processor. The flexibility of this new structure will permit both 'block' and 'sample' filter operations and is therefore proposed for the implementation of wavelet filters and generalized multirate filter banks. The results are equally valid for other array forms (such as unconstrained nonlinear arrays), for space-based Multiple Beam Antennas (MBAs) in satellite communications, and for other applications such as radar and sonar adaptive nulling.<>