Intelligent Reflecting Surface Aided MISO Uplink Communication Network: Feasibility and Power Minimization for Perfect and Imperfect CSI

Yang Liu, Jun Zhao, Ming Li, Qingqing Wu

Abstract

In this paper, we consider the weighted sum-power minimization under quality-of-service (QoS) constraints in the multi-user multi-input-single-output (MISO) uplink wireless network assisted by intelligent reflecting surface (IRS). We perform a comprehensive investigation on various aspects of this problem. First, when users have sufficient transmit powers, we present a new sufficient condition guaranteeing arbitrary information rate constraints. This result strengthens the feasibility condition in existing literature. Then, we design novel penalty dual decomposition (PDD) based and nonlinear equality constrained alternative direction method of multipliers (neADMM) based solutions to tackle the IRS-dependent-QoS-constraints, which effectively solve the feasibility check and power minimization problems. Besides, we further extend our proposals to the cases where channel status information (CSI) is imperfect and develop an online stochastic algorithm that satisfy QoS constraints stochastically without requiring prior knowledge of CSI errors. Extensive numerical results are presented to verify the effectiveness of our proposed algorithms.

Introduction

In modern multi-user wireless networks, mobile devices often struggle to maintain high data rates (Quality-of-Service, or QoS) while conserving battery life. Increasing transmit power to meet these rates leads to faster battery drain and heightened inter-user interference. Intelligent Reflecting Surfaces (IRS) have emerged as a transformative technology to address this. By using metasurfaces to "steer" radio waves through phase adjustments, IRS can significantly improve the signal environment without the power consumption of active amplifiers.

This paper addresses the challenge of minimizing total transmit power in an IRS-assisted MISO uplink network while strictly adhering to QoS targets. The optimization problem is inherently difficult due to the non-convex "unit-modulus" constraints of IRS elements and the complexities of multi-user interference.

System Model and Feasibility

The study considers a multi-user MISO uplink where multiple single-antenna users (MUs) communicate with a multi-antenna Access Point (AP), aided by an IRS.

Figure 1: System model of an IRS-assisted multi-user MISO uplink network.

The network's performance depends on the direct links ($h_{d,k}$) and the reflected links through the IRS ($G$ and $h_{r,k}$). A critical first step is determining feasibility: can the IRS configuration and user power levels actually meet the required data rates? The authors derive a new sufficient condition that provides a stronger theoretical bound than previous literature, ensuring that rate constraints are achievable under given channel conditions.

Proposed Algorithmic Frameworks

To solve the non-convex power minimization problem under perfect Channel State Information (CSI), the authors propose two main frameworks:

1. Penalty Dual Decomposition (PDD)

The PDD-based method decomposes the complex global problem into smaller, more manageable sub-problems. By introducing auxiliary variables and applying a penalty for discrepancies between these variables, the algorithm iteratively approaches an optimal solution that respects the IRS's physical constraints.

2. Nonlinear Equality ADMM (neADMM)

The neADMM framework is a modern optimization approach designed to handle the tricky unit-modulus constraints (where each reflecting element only changes phase, not amplitude) more efficiently than traditional methods like Semidefinite Relaxation (SDR).

Figure 4: The neADMM framework designed for efficient handling of nonlinear equality constraints in IRS optimization.

Handling Imperfect CSI

Real-world deployments rarely have access to perfect CSI. To account for measurement noise and estimation errors, the authors develop an online stochastic algorithm. Unlike robust optimization methods that require prior knowledge of error distributions, this algorithm "learns" and adapts to a stream of noisy estimates, ensuring that QoS constraints are satisfied stochastically (on average) over time.

Key Insights and Results

The research yields several significant findings:

  • Limitations of SDR: Traditional Semidefinite Relaxation (SDR) often fails in IRS phase optimization because the randomization step required to enforce unit-modulus constraints can destroy the solution's optimal structure.
  • Convergence and Precision: The proposed neADMM algorithm demonstrates superior convergence, reaching high precision ($10^{-3}$ to $10^{-4}$) within a few dozen iterations, outperforming standard ADMM and CCP benchmarks.
  • Expanded Feasibility: The inclusion of an IRS dramatically widens the network's operational range. For instance, at high rate requirements, a system without IRS might be feasible only 5% of the time, while a 120-element IRS boosts this to nearly 40%.
  • Power Efficiency: Optimized IRS configurations significantly reduce total transmit power by effectively "focusing" user energy toward the AP, creating a "SINR margin" that allows for substantial energy savings.

Conclusion

This work provides a comprehensive toolkit for optimizing IRS-assisted uplink networks. By combining a new feasibility analysis with high-performance PDD and neADMM algorithms, the authors offer robust solutions for both perfect and imperfect CSI scenarios. Future research may explore hardware limitations like discrete phase shifters and the overhead associated with channel estimation for large-scale surfaces.

Summary updated

Abstract

In this paper, we consider the weighted sum-power minimization under quality-of-service (QoS) constraints in the multi-user multi-input-single-output (MISO) uplink wireless network assisted by intelligent reflecting surface (IRS). We perform a comprehensive investigation on various aspects of this problem. First, when users have sufficient transmit powers, we present a new sufficient condition guaranteeing arbitrary information rate constraints. This result strengthens the feasibility condition in existing literature. Then, we design novel penalty dual decomposition (PDD) based and nonlinear equality constrained alternative direction method of multipliers (neADMM) based solutions to tackle the IRS-dependent-QoS-constraints, which effectively solve the feasibility check and power minimization problems. Besides, we further extend our proposals to the cases where channel status information (CSI) is imperfect and develop an online stochastic algorithm that satisfy QoS constraints stochastically without requiring prior knowledge of CSI errors. Extensive numerical results are presented to verify the effectiveness of our proposed algorithms.
BibTeX
@article{liu2020intelligent,
  author = {Yang Liu and Jun Zhao and Ming Li and Qingqing Wu},
  title = {{Intelligent Reflecting Surface Aided MISO Uplink Communication Network: Feasibility and Power Minimization for Perfect and Imperfect CSI}},
  journal = {IEEE Transactions on Communications},
  year = {2020},
  doi = {10.1109/tcomm.2020.3040404},
}
Figure