Coordinate Update Algorithms for Robust Power Loading for the MU-MISO Downlink with Outage Constraints

TL;DR

This paper introduces coordinate update algorithms for robust power loading in MU-MISO downlink systems with outage constraints, achieving optimal or near-optimal solutions with lower computational complexity.

Contribution

It develops a coordinate descent algorithm for a non-convex deterministic reformulation of outage-constrained power allocation, ensuring convergence to a global optimum from feasible starting points.

Findings

Algorithms outperform existing robust power loading methods.

Proposed methods achieve better power efficiency under outage constraints.

Lower computational cost compared to traditional approaches.

Abstract

We consider the problem of power allocation for the single-cell multi-user (MU) multiple-input single-output (MISO) downlink with quality-of-service (QoS) constraints. The base station acquires an estimate of the channels and, for a given beamforming structure, designs the power allocation so as to minimize the total transmission power required to ensure that target signal-to-interference-and-noise ratios at the receivers are met, subject to a specified outage probability. We consider scenarios in which the errors in the base station's channel estimates can be modelled as being zero-mean and Gaussian. Such a model is particularly suitable for time division duplex (TDD) systems with quasi-static channels, in which the base station estimates the channel during the uplink phase. Under that model, we employ a precise deterministic characterization of the outage probability to transform the chance-constrained formulation to a deterministic one. Although that deterministic formulation is not convex, we develop a coordinate descent algorithm that can be shown to converge to a globally optimal solution when the starting point is feasible. Insight into the structure of the deterministic formulation yields approximations that result in coordinate update algorithms with good performance and significantly lower computational cost. The proposed algorithms provide better performance than existing robust power loading algorithms that are based on tractable conservative approximations, and can even provide better performance than robust precoding algorithms based on such approximations.