Joint Power Allocation and Reflecting-Element Activation for Energy Efficiency Maximization in IRS-Aided Communications Under CSI Uncertainty

TL;DR

This paper proposes robust algorithms for joint power allocation and reflecting-element activation in IRS-assisted communication systems to maximize energy efficiency under CSI uncertainty, balancing optimality and computational complexity.

Contribution

It introduces two algorithms, an AO-based suboptimal method and a B&B approach, for solving a mixed-integer robust optimization problem without external solvers.

Findings

Proposed algorithms outperform baseline schemes.

AO achieves near-optimal performance in most cases.

B&B algorithm guarantees global optimality with low average complexity.

Abstract

We study the joint power allocation and reflecting element (RE) activation to maximize the energy efficiency (EE) in communication systems assisted by an intelligent reflecting surface (IRS), taking into account imperfections in channel state information (CSI). The robust optimization problem is mixed integer, i.e., the optimization variables are continuous (transmit power) and discrete (binary states of REs). In order to solve this challenging problem we develop two algorithms. The first one is an alternating optimization (AO) method that attains a suboptimal solution with low complexity, based on the Lambert W function and a dynamic programming (DP) algorithm. The second one is a branch-and-bound (B&B) method that uses AO as its subroutine and is formally guaranteed to achieve a globally optimal solution. Both algorithms do not require any external optimization solver for their implementation. Furthermore, numerical results show that the proposed algorithms outperform the baseline schemes, AO achieves near-optimal performance in most cases, and B&B has low computational complexity on average.