Técnicas de Inteligência Computacional Aplicadas à Estabilidade nos Sistemas DSL

Abstract

The stability of digital subscriber lines (DSL) becomes a major challenge when tripleplay services are provisioned. These services must be reliable, while requiring relatively high transmission rates and low latency for an overall improved quality of experience. However, the stabilization tools available in DSL systems such as the automatic margin adaptation (AMA), interleaving, impulse noise protection (INP) and virtual noise (VN) usually trade o bit rate or latency in an attempt of improving the stability. The current work, on the other hand, presents a novel approach to the stabilization of the DSL lines. This approach is based on multiuser spectra coordination that seeks to mitigate crosstalk interference e ect such that feasible bit rate requirements and power restrictions are met with optimized noise margin. As only physical layer parameters are adjusted, latency is not a ected by the method. The problem is posed as a multiuser margin optimization problem and solved by evolutionary algorithms in conjunction with power minimization techniques. Simulations show that the proposed method provides signi cant improvements in terms of noise margin when compared to a previous algorithm, achieving higher protection for the network without penalizing bit rate or latency.