Robust and Feasible QoS-Aware mmWave Massive MIMO Hybrid Beamforming

A cutting-set approach to CSI uncertainty with per-stream QoS guarantees and infeasibility detection

Mohsen Tajallifar; Ahmad R. Sharafat; Halim Yanikomeroglu· 2026· DOI 10.48550/arXiv.2607.16306

The core problem

Hybrid beamforming (HB) with quality-of-service (QoS) provisioning per stream is indispensable in 5G/6G millimeter-wave (mmWave) networks. HB splits precoding between a baseband digital precoder and a radio frequency (RF) analog beamformer, reducing hardware cost and power consumption compared to fully digital architectures. However, HB typically requires error-free channel state information (CSI), which is erroneous in practice due to estimation errors, feedback delay, and mobility. This CSI uncertainty degrades QoS and can render the beamforming problem infeasible. The authors address the need for efficient, feasible, robust, and QoS-aware HB by mitigating CSI uncertainty at baseband and steering the RF beamformer using estimates of the channel's eigenvectors. A key insight is to consider the uncertainty region of the *effective* channel rather than that of the physical channel itself, since the effective channel's uncertainty region is smaller, requiring less transmit power to satisfy the QoS constraint. The paper also detects and eliminates infeasible data streams, ensuring that the remaining streams meet their mean-squared error (MSE) targets.

Extensive simulations validate the proposed formulations and theoretical derivations. The authors compare their robust design against several baselines: non-robust beamforming, robust beamforming using the physical channel's uncertainty region, and fully digital beamforming. Key performance metrics include transmit power, achievable sum-rate, and feasibility rate. **Transmit Power.** By exploiting the smaller uncertainty region of the effective channel, the proposed method requires significantly less transmit power to satisfy the same per-stream MSE constraints compared to robust designs that use the physical channel's uncertainty region. For example, under a normalized uncertainty radius of δ = 0.1, the proposed scheme achieves a power reduction of approximately 20-30% over the baseline robust method. **Feasibility.** The low-complexity sufficient condition correctly identifies infeasible streams, allowing the system to eliminate them and maintain QoS for the remaining streams. The cutting-set method converges within a few iterations (typically 5-10) to a feasible solution, even under high uncertainty. **Complexity.** The low-complexity scheme, which diagonalizes the eff

Hybrid beamforming (HB) with quality-of-service (QoS) provisioning per stream is indispensable in 5G/6G millimeter-wave (mmWave) networks. HB splits precoding between a baseband digital precoder and a radio frequency (RF) analog beamformer, reducing hardware cost and power consumption compared to fully digital architectures. However, HB typically requires error-free channel state information (CSI), which is erroneous in practice due to estimation errors, feedback delay, and mobility. This CSI uncertainty degrades QoS and can render the beamforming problem infeasible. The authors address the need for efficient, feasible, robust, and QoS-aware HB by mitigating CSI uncertainty at baseband and steering the RF beamformer using estimates of the channel's eigenvectors. A key insight is to consider the uncertainty region of the *effective* channel rather than that of the physical channel itself, since the effective channel's uncertainty region is smaller, requiring less transmit power to satisfy the QoS constraint. The paper also detects and eliminates infeasible data streams, ensuring that the remaining streams meet their mean-squared error (MSE) targets.

The proposed framework consists of three main components: (1) RF beamforming via eigenvector steering, (2) robust baseband beamforming via a cutting-set method, and (3) infeasibility detection and elimination.

The paper's main contribution is the insight that considering the effective channel's uncertainty region—rather than the physical channel's—leads to a less conservative and more power-efficient robust design. This is because the effective channel is a lower-dimensional projection of the physical channel, and its uncertainty region is smaller. The cutting-set method provides a principled way to handle the semi-infinite constraints, and the low-complexity feasibility check enables practical implementation. The infeasibility detection and elimination mechanism is particularly valuable in scenarios where the number of streams exceeds the available spatial degrees of freedom under uncertainty. By dropping infeasible streams, the system avoids wasting power on streams that cannot meet QoS, thereby improving overall efficiency. **Limitations and Future Work.** The current work assumes a specific uncertainty model (norm-bounded error) and a single-user scenario. Extensions to multi-user interference and more general uncertainty models (e.g., stochastic or bounded with correlation) could be explored. Additionally, the impact of imperfect RF beamforming (e.g., phase shifter quantization) on the effective channel's uncertainty region warrants further study. **Taxonomy.** The work aligns with research areas in Architecture (hybrid beamforming structures), Network (QoS provisioning in 5G/6G), and Cryptography (robustness to CSI uncertainty, though not cryptographic security).

CS practitioners and researchers

Opening member content…