[Paper Review] Terahertz-Band Joint Ultra-Massive MIMO Radar-Communications: Model-Based and Model-Free Hybrid Beamforming

Ahmet M. Elbir, Kumar Vijay Mishra+1 more|arXiv (Cornell University)|Feb 27, 2021

Millimeter-Wave Propagation and Modeling80 references169 citations

TL;DR

This paper proposes a hybrid beamforming framework for terahertz-band joint radar-communications using an ultra-massive MIMO system with a group-of-subarrays (GoSA) architecture. It introduces model-based and model-free deep learning techniques to reduce hardware complexity and computation time, while mitigating beam split through phase correction, achieving high spectral efficiency and robust radar beampatterns with 500× lower computation time than conventional methods.

ABSTRACT

Wireless communications and sensing at terahertz (THz) band are increasingly investigated as promising short-range technologies because of the availability of high operational bandwidth at THz. In order to address the extremely high attenuation at THz, ultra-massive multiple-input multiple-output (MIMO) antenna systems have been proposed for THz communications to compensate propagation losses. However, the cost and power associated with fully digital beamformers of these huge antenna arrays are prohibitive. In this paper, we develop wideband hybrid beamformers based on both model-based and model-free techniques for a new group-of-subarrays (GoSA) ultra-massive MIMO structure in low-THz band. Further, driven by the recent developments to save the spectrum, we propose beamformers for a joint ultra-massive MIMO radar-communications system, wherein the base station serves multi-antenna user equipment (RX), and tracks radar targets by generating multiple beams toward both RX and the targets. We formulate the GoSA beamformer design as an optimization problem to provide a trade-off between the unconstrained communications beamformers and the desired radar beamformers. To mitigate the beam split effect at THz band arising from frequency-independent analog beamformers, we propose a phase correction technique to align the beams of multiple subcarriers toward a single physical direction. To further decrease the ultra-massive MIMO computational complexity and enhance robustness, we also implement deep learning solutions to the proposed model-based hybrid beamformers. Numerical experiments demonstrate that both techniques outperform the conventional approaches in terms of spectral efficiency and radar beampatterns, as well as exhibiting less hardware cost and computation time.

Motivation & Objective

• Address the high path loss and hardware cost in terahertz-band ultra-massive MIMO systems for joint radar-communications.
• Develop hybrid beamforming techniques that balance spectral efficiency and hardware complexity in low-THz band.
• Mitigate beam split caused by frequency-independent analog beamformers in wideband THz systems.
• Reduce channel feedback overhead by exploiting second-order channel statistics.
• Enable low-complexity, high-performance beamforming using deep learning for ultra-massive MIMO systems.

Proposed method

• Propose a GoSA ultra-massive MIMO structure to reduce phase shifter count and hardware cost compared to fully connected arrays.
• Formulate hybrid beamforming as an optimization problem trading off communication and radar performance.
• Implement a phase correction technique to align beams across subcarriers by adjusting analog beamformer phases using the central frequency fc.
• Use channel state information (CSI) and channel covariance matrices to design model-based beamformers with reduced feedback overhead.
• Apply deep learning to model-free beamforming, training on large datasets to approximate optimal beamformers with parallel GPU processing.
• Integrate beam split correction into the baseband beamformer to avoid costly phase shifter reconfiguration.

Experimental results

Research questions

• RQ1 How can hybrid beamforming be designed for ultra-massive MIMO in joint radar-communications systems to balance spectral efficiency and hardware cost?
• RQ2 What is the impact of frequency-independent analog beamformers on beam alignment in wideband THz systems, and how can beam split be corrected?
• RQ3 Can model-free deep learning-based beamforming achieve performance close to CSI-based methods with significantly reduced computation time?
• RQ4 How does exploiting second-order channel statistics reduce feedback overhead in ultra-massive MIMO systems?
• RQ5 What is the performance trade-off between fully connected and partially connected GoSA structures in terms of spectral efficiency and radar accuracy?

Key findings

• The GoSA structure reduces hardware complexity by minimizing the number of phase shifters compared to fully connected and AoSA architectures.
• The proposed beam split correction technique restores spectral efficiency in wideband systems, eliminating performance degradation from frequency-independent analog beamformers.
• Model-free deep learning beamforming achieves spectral efficiency within 5% of CSI-based methods while reducing computation time by approximately 500 times.
• The channel covariance matrix-based beamformer reduces feedback overhead by 90% compared to CSI-based methods, with only a minor performance loss.
• Partially connected GoSA structures show slightly worse performance than fully connected ones but offer a favorable trade-off between complexity and performance.
• The model-free approach achieves a computation time of 0.0058 seconds, compared to 2.124 seconds for the MO-based method, demonstrating superior scalability for ultra-massive MIMO systems.