Trade-offs in Reliability and Performance Using Selective Beamforming for Ultra-Massive MIMO

Technical Summary: Trade-offs in Reliability and Performance Using Selective Beamforming for Ultra-Massive MIMO

Problem Statement
This paper addresses the optimization challenges inherent in Ultra-Massive MIMO (UM-MIMO) communication systems, specifically focusing on array selection and beamforming within dynamic operational contexts. Traditional beamforming designs often overlook the varying reliability or structural health of individual antenna elements, potentially allocating power to compromised components and leading to suboptimal performance or system vulnerability. Furthermore, modern networks face heterogeneous user demands (e.g., IoT vs. video streaming) and strict Quality of Service (QoS) requirements, necessitating resource allocation strategies that go beyond uniform distribution. The core problem is to jointly optimize beamforming and sparse array selection to maximize system performance while incorporating antenna health data and ensuring proportional fairness among users, all within a constrained, non-convex, and non-smooth optimization landscape.

Methodology
The authors propose a novel framework that integrates antenna health information directly into the beamforming optimization process. The system model considers a mmWave node with $N_t$ transmit antennas serving $M$ single-antenna users. The received signal is modeled with additive white Gaussian noise, and the Signal-to-Interference-plus-Noise Ratio (SINR) is calculated for each user.

The core of the methodology involves formulating a constrained optimization problem (Equation 5) that maximizes a weighted sum of user rates while penalizing the use of unreliable antenna elements. This is achieved by introducing a reliability matrix $\beta$, where elements range from 0 (failure) to 1 (full operation). A sparsity-promoting regularization term, weighted by a parameter $\gamma$ (or $\rho_s$ in the formulation), is added to the objective function. This term encourages the selection of reliable elements and suppresses connections to unreliable ones, resulting in a sparse beamforming matrix.

To solve this non-convex and non-smooth problem, the authors employ a Proximal-Gradient Dual Ascent (PGDA) algorithm (Algorithm 1). This approach utilizes dual variables to handle constraints such as total transmit power ($P_t$) and minimum rate requirements ($R_{min}$). The algorithm iteratively updates the beamforming weights using gradient ascent and applies a proximal operator involving soft-thresholding to enforce sparsity based on the reliability mask. The method avoids unnecessary mathematical relaxations (such as Semidefinite Relaxation) often found in related works, aiming for faster convergence and better adaptability.

Key Contributions
The paper outlines three primary contributions:

1. Antenna-Health Aware Selection: The introduction of a new array selection criterion that incorporates real-time antenna health information to correct beam patterns and enhance system reliability.
2. Optimization Technique: The application of lightweight optimization techniques, specifically gradient dual ascent and dual proximal-gradient ascent, to directly tackle the non-convex challenges of sparse array selection without relying on heavy relaxation methods.
3. Proportional Fairness: The implementation of a fairness mechanism that aligns resource allocation with system constraints while guaranteeing minimum rate requirements for all users, ensuring equitable distribution in heterogeneous network environments.

Experimental Results
The proposed solution was evaluated through extensive simulations using a 28 GHz mmWave band with 64 transmit antennas and 4 users. The system operated with a total bandwidth of 3 GHz and a transmit power of 2 kW. The study varied the sparsity promotion weight ($\gamma$) to observe its impact on performance metrics.

Key findings from the simulations include:

• Trade-off between Efficiency and Reliability: As the sparsity weight $\gamma$ increases, Spectral Efficiency (SE) and communication rates ($R_i$) decrease. For instance, SE dropped from 2.32 bps/Hz at $\gamma=0$ to 1.78 bps/Hz at $\gamma=333.4$.
• Sparsity and Density: Higher $\gamma$ values significantly reduced the Beamforming Matrix Density (BMD), moving from 100% density to approximately 64% density, indicating a shift toward sparse array usage.
• Reliability Improvement: Conversely, Antenna Reliability ($R_L$) increased dramatically with higher $\gamma$, rising from 0% to nearly 93%. This demonstrates that the system successfully prioritizes redundant, reliable elements over the entire array.
• Power Consumption: Power consumption showed marginal variation, slightly increasing as sparsity increased, suggesting that the efficiency gains in terms of reliability do not come at a prohibitive power cost in this specific configuration.

Significance and Claims
The paper claims that its work advances the field of UM-MIMO by providing innovative solutions for array selection and beamforming that explicitly account for structural health. The authors assert that their approach establishes a foundation for future research by demonstrating how to balance the critical trade-off between high performance (spectral efficiency) and system robustness (reliability).

The significance of the work lies in its ability to create "self-aware" communication systems that can dynamically recalibrate power distribution based on the physical condition of the antenna array. This is presented as essential for the practical deployment of UM-MIMO systems, particularly in scenarios requiring uninterrupted operation, such as critical infrastructure or defense applications. The authors conclude that while increasing sparsity reduces raw throughput, it significantly enhances fault tolerance and operational integrity, offering a viable strategy for resilient 5G and 6G network design. The study modestly notes that the current 64-antenna simulation was chosen for computational feasibility on desktop hardware, with future work intended to explore scalability to larger arrays.