On Numerical Robustness of Constrained RLS-Like Algorithms

Antônio L. L. Ramos, José A. Apolinário Jr, Marcello L. R. de Campos

DOI: 10.14209/sbrt.2004.292

Evento: XXI Simpósio Brasileiro de Telecomunicações (SBrT2004)

Keywords:

Abstract

A number of Constrained RLS–type algo rithms have been recently proposed for linearly con strained adaptive applications such as adaptive beamform ing. Some of these algorithms, alternatives to employing the Generalized Sidelobe Canceller structure, claim to be robust whether for using a special correction term or for employing numerically stable rotations. Nevertheless, most of these algorithms diverge in long–run simulations or are too sensitive to changes in the forgetting factor. This paper deals with a performance comparison of many Constrained RLS-like algorithms—including two other fast converg ing non-RLS algorithms—and their GSC counterparts in terms of numerical stability, speed of convergence, and computational complexity. An efficient and recently proposed Householder structure is also taken into account such that the algorithms presenting the most promising results are summarized and appropriate options for fast converging adaptive beamforming applications accrue.

Download