Terahertz-Band Joint Ultra-Massive MIMO Radar-Communications:\n Model-Based and Model-Free Hybrid Beamforming

Abstract

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