Multi-RIS Assisted Hybrid Beamforming Design for Terahertz Massive MIMO Systems

Abstract

This paper presents an innovative approach for terahertz (THz) band communications, utilizing reconfigurable intelligent surfaces (RISs) to implement an angular-based hybrid beamforming (AB-HBF). This study examines two significant THz channel scenarios with cost-efficient solutions utilizing RISs to enhance performance. The first scenario ensures reliable communication in the presence of obstacles blocking the direct path, where RISs offer an alternative transmission path. The second scenario leverages multiple-RIS in line-of-sight dominant or sparse environments to increase the rank of the channel matrix, consequently improving the achievable rate. By introducing an innovative HBF design for multiple-RIS-assisted THz massive MIMO systems, a three-stage design is proposed based on a geometry-based THz channel model. This design encompasses the transmit and receive radio frequency (RF) beamformers, transmit/receive baseband (BB) precoder/combiner, and RIS phase shift matrix. A particle swarm optimization-based solution is employed to design the RIS phase reflections. This design methodology is extended to the multiple-RIS-aided scenario, optimizing the phase shift matrices of each RIS. We also provide sub-connected AB-HBF configuration by offering a remarkable reduction in the total number of phase shifters compared to the fully-connected AB-HBF. Additionally, a deep learning-based phase shift design is incorporated to effectively optimize the RIS configurations, significantly reducing the computational time required for system calibration in dynamic THz environments. Extensive numerical experiments demonstrate the significant enhancement in the achievable rate of THz AB-HBF systems by incorporating RISs, mitigating the wireless propagation disruptions as well as reducing hardware cost/complexity and power consumption for massive MIMO systems.

Keywords

• Hybrid beamforming
• terahertz band
• massive MIMO
• low CSI overhead
• reconfigurable intelligent surface