Research on Large Adaptive Arrays.

Most adaptive filters are designed to achieve minimization of estimation error. Other design goals could include satisfaction of constraints as well as minimization of mean square error. The paper proposes a particular set of performance criteria, involving 'soft constraints'. A 'soft-constraint LMS algorithm' is derived from a performance function which is the sum of mean square error and weighted squared constraint violation errors. The soft constraint algorithm is applied to adaptive antenna arrays as an example, which includes a demonstration of the effects of varying the softness of the constraints. Also, convergence properties of the algorithm are presented. A relation between the output power of a signal from a converged soft constraint LMS adaptive filter and the signal's input power is derived, which demonstrates some unexpected behavior. (Author)
Prepared in cooperation with Stanford Univ., CA under contract N00019-78-C-0276.