Joint power control and user scheduling in multicell wireless networks: Capacity scaling laws

TL;DR

This paper investigates the capacity scaling laws in multicell wireless networks with joint power control and user scheduling, showing that distributed algorithms are feasible and interference impact diminishes as the number of users grows large.

Contribution

It introduces asymptotic analysis of capacity scaling laws using extreme value theory, demonstrating the feasibility of distributed algorithms in large multicell networks.

Findings

Distributed algorithms are obtainable in large user regimes.

Capacity scaling is dominated by path loss or fading depending on scenarios.

Interference impact on capacity diminishes asymptotically.

Abstract

We address the optimization of the sum rate performance in multicell interference-limited singlehop networks where access points are allowed to cooperate in terms of joint resource allocation. The resource allocation policies considered here combine power control and user scheduling. Although very promising from a conceptual point of view, the optimization of the sum of per-link rates hinges, in principle, on tough issues such as computational complexity and the requirement for heavy receiver-to-transmitter channel information feedback across all network cells. In this paper, we show that, in fact, distributed algorithms are actually obtainable in the asymptotic regime where the numbers of users per cell is allowed to grow large. Additionally, using extreme value theory, we provide scaling laws for upper and lower bounds for the network capacity (sum of single user rates over all cells), corresponding to zero-interference and worst-case interference scenarios. We show that the scaling is either dominated by path loss statistics or by small-scale fading, depending on the regime and user location scenario. We show that upper and lower rate bounds behave in fact identically, asymptotically. This remarkable result suggests not only that distributed resource allocation is practically possible but also that the impact of multicell interference on the capacity (in terms of scaling) actually vanishes asymptotically.