| The Digital Subscriber Line (DSL) environment is characterized by highly frequency-selective attenuation and by potentially large crosstalk between users. DSL resource allocation algorithms allow for an efficient use of the DSL network by managing the crosstalk-induced interference while also taking the frequency-selectivity into account. Previous resource allocation algorithms were based on worst-case situations. The more recent Dynamic Spectrum Management (DSM) resource allocation algorithms are able to constantly adapt to the channel characteristics. At one end of the DSM algorithms, there is the low-performing user-level Iterative Water-Filling (IWF) algorithm with low computational complexity and full distributivity. At the other end, the better-performing network-level algorithms are not fully distributable and have higher computational complexities.;The Autonomous Spectrum Balancing using Multiple Reference Lines (ASB-MRL) algorithm is then presented as an algorithm capable of bridging the performance gap between the fully-distributable and low-complexity IWF, and the high-performing network-level algorithms while maintaining the benefits of each. Following that, a set of conditions on the virtual network formed by the multiple reference lines is produced to ensure that ASB-MRL allocates the resources in a near-optimal manner.;In this thesis, an overview of the DSL environment and its resource allocation algorithms is presented. Then, three alternatives with a smaller computational load than the bisection method used in IWF's water-filling sub-algorithm are presented. These alternatives include a novel projection method for a specific case and the novel Generalized Recursive Water-Filling (GRWF) algorithm for the generalized resource allocation problem. |
