Fair dynamic spectrum management for ordered zero forcing nonlinear precoder based G.fast transmission

In the G.fast frequency range with strong levels of crosstalk, nonlinear precoding (NLP) is shown to be a near-optimal technique for crosstalk cancellation. In contrast, current methods for multi-tone NLP user encoding ordering (UEO) aiming for some fairness between users suffer from substantial suboptimality. In this work, we develop a novel dynamic spectrum management algorithm for ordered zeroforcing NLP that enforces an alpha fairness policy. Since finding the optimal UEO is an optimization problem with exploding combinatorial growth, the proposed algorithm uses an low-complex iterative method which provides near-optimal approximate solutions. Simulations demonstrate the novel algorithm achieving a trade-off between fairness and performance that outperforms current UEO methods.