Convergence of Iterative Water-Filling in Multi-User Non-Cooperative Power Control: A Comprehensive Analysis for Sequential, Simultaneous, and Asynchronous Schemes

TL;DR

This paper provides a comprehensive analysis of the convergence properties of iterative water-filling algorithms in multi-user wireless networks under various update schemes, offering practical guidelines for ensuring stability and efficiency.

Contribution

It introduces unified convergence conditions for sequential, simultaneous, and asynchronous water-filling algorithms using contraction mapping principles, including robustness considerations.

Findings

Convergence guaranteed under spectral radius and diagonal dominance conditions.

Asynchronous schemes can achieve convergence with proper step-size control.

Robustness to noise and partial updates is demonstrated.

Abstract

Non-cooperative game theory provides a robust framework for analyzing distributed resource allocation in multi-user wireless networks, with \emph{Iterative Water-Filling} (IWF) emerging as a canonical solution for power control problems. Although classical fixed-point theorems guarantee the existence of a Nash Equilibrium (NE) under mild concavity and compactness conditions, the convergence of practical iterative algorithms to that equilibrium remains a challenging endeavor. This challenge intensifies under varying update schedules, interference regimes, and imperfections such as channel estimation errors or feedback delay. In this paper, we present an in-depth examination of IWF in multi-user systems under three different update schemes: (1) synchronous \emph{sequential} updates, (2) synchronous \emph{simultaneous} updates, and (3) \emph{totally asynchronous} updates. We first formulate the water-filling operator in a multi-carrier environment, then recast the iterative process as a fixed-point problem. Using contraction mapping principles, we demonstrate sufficient conditions under which IWF converges to a unique NE and highlight how spectral radius constraints, diagonal dominance, and careful step-size selection are pivotal for guaranteeing convergence. We further discuss robustness to measurement noise, partial updates, and network scaling to emphasize the practical viability of these schemes. This comprehensive analysis unifies diverse threads in the literature while offering novel insights into asynchronous implementations. Our findings enable network designers to ascertain system parameters that foster both stable convergence and efficient spectrum usage.