Distributed power allocation and user assignment in OFDMA cellular networks

We consider a wireless OFDMA cellular network and address the problem of jointly allocating power to frequencies (subbands) and assigning users to cells, for a variety of elastic and inelastic services. The goal is to maximize the sum of the users´ throughput utility functions subject to various constraints, such as minimum throughput requirements. The problem naturally fits into a Network Utility Maximization (NUM) framework with a mixture of concave (e.g., data rates) and nonconcave (e.g., voice/video streaming) utilities. The hardness of this nonconvex, mixed integer program prohibits the use of standard convex optimization algorithms, or efficient combinatorial approximation techniques. We devise a randomized algorithm for the said NUM problem, whose proof of asymptotic optimality is derived from the classical framework of interacting particle systems, via a judiciously selected neighborhood structure. The proposed algorithm is highly distributed, asynchronous, requires limited computational effort per node/iteration, and yields provable convergence in the limit. Several numerical experiments are presented to illustrate the convergence speed and performance of the proposed method.