Deep Learning–Based Channel Extrapolation and Multiuser Beamforming for RIS-aided Terahertz Massive MIMO Systems over Hybrid-Field Channels
The reconfigurable intelligent surface (RIS) is a promising technology for terahertz (THz) massive multiple-input multiple-output (MIMO) communication systems. However, acquiring high-dimensional channel state information (CSI) and realizing efficient active/passive beamforming for RIS are challengi...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
American Association for the Advancement of Science (AAAS)
2024-01-01
|
| Series: | Intelligent Computing |
| Online Access: | https://spj.science.org/doi/10.34133/icomputing.0065 |
| _version_ | 1797258746909425664 |
|---|---|
| author | Yang Wang Zhen Gao Sheng Chen Chun Hu Dezhi Zheng |
| author_facet | Yang Wang Zhen Gao Sheng Chen Chun Hu Dezhi Zheng |
| author_sort | Yang Wang |
| collection | DOAJ |
| description | The reconfigurable intelligent surface (RIS) is a promising technology for terahertz (THz) massive multiple-input multiple-output (MIMO) communication systems. However, acquiring high-dimensional channel state information (CSI) and realizing efficient active/passive beamforming for RIS are challenging owing to its cascaded channel structure and lack of signal processing units. To overcome these challenges, this study proposes a deep learning (DL)-based physical signal processing scheme for RIS-aided THz massive MIMO systems over hybrid far-near field channels wherein channel estimation with low pilot overhead and robust beamforming are implemented. Specifically, first, an end-to-end DL-based channel estimation framework that consists of pilot design, CSI feedback, subchannel estimation, and channel extrapolation is introduced. In this framework, only some RIS elements are first activated, a subsampling RIS channel is then estimated, and a DL-based extrapolation network is finally used to reconstruct the full-dimensional CSI. Next, to maximize the sum rate under imperfect CSI, a DL-based scheme is developed to simultaneously design hybrid active beamforming at the base station and passive beamforming at the RIS. Simulation results show that the proposed channel extrapolation scheme achieves better CSI reconstruction performance than conventional schemes while greatly reducing pilot overhead. Moreover, the proposed beamforming scheme outperforms conventional schemes in terms of robustness to imperfect CSI. |
| first_indexed | 2024-04-24T22:58:26Z |
| format | Article |
| id | doaj.art-abe1c6e9d87d462bba8ef68349db43ed |
| institution | Directory Open Access Journal |
| issn | 2771-5892 |
| language | English |
| last_indexed | 2024-04-24T22:58:26Z |
| publishDate | 2024-01-01 |
| publisher | American Association for the Advancement of Science (AAAS) |
| record_format | Article |
| series | Intelligent Computing |
| spelling | doaj.art-abe1c6e9d87d462bba8ef68349db43ed2024-03-18T02:30:56ZengAmerican Association for the Advancement of Science (AAAS)Intelligent Computing2771-58922024-01-01310.34133/icomputing.0065Deep Learning–Based Channel Extrapolation and Multiuser Beamforming for RIS-aided Terahertz Massive MIMO Systems over Hybrid-Field ChannelsYang Wang0Zhen Gao1Sheng Chen2Chun Hu3Dezhi Zheng4MIIT Key Laboratory of Complex-Field Intelligent Sensing, Beijing Institute of Technology, Beijing, China.MIIT Key Laboratory of Complex-Field Intelligent Sensing, Beijing Institute of Technology, Beijing, China.School of Electronics and Computer Science, University of Southampton, Southampton, UK.MIIT Key Laboratory of Complex-Field Intelligent Sensing, Beijing Institute of Technology, Beijing, China.MIIT Key Laboratory of Complex-Field Intelligent Sensing, Beijing Institute of Technology, Beijing, China.The reconfigurable intelligent surface (RIS) is a promising technology for terahertz (THz) massive multiple-input multiple-output (MIMO) communication systems. However, acquiring high-dimensional channel state information (CSI) and realizing efficient active/passive beamforming for RIS are challenging owing to its cascaded channel structure and lack of signal processing units. To overcome these challenges, this study proposes a deep learning (DL)-based physical signal processing scheme for RIS-aided THz massive MIMO systems over hybrid far-near field channels wherein channel estimation with low pilot overhead and robust beamforming are implemented. Specifically, first, an end-to-end DL-based channel estimation framework that consists of pilot design, CSI feedback, subchannel estimation, and channel extrapolation is introduced. In this framework, only some RIS elements are first activated, a subsampling RIS channel is then estimated, and a DL-based extrapolation network is finally used to reconstruct the full-dimensional CSI. Next, to maximize the sum rate under imperfect CSI, a DL-based scheme is developed to simultaneously design hybrid active beamforming at the base station and passive beamforming at the RIS. Simulation results show that the proposed channel extrapolation scheme achieves better CSI reconstruction performance than conventional schemes while greatly reducing pilot overhead. Moreover, the proposed beamforming scheme outperforms conventional schemes in terms of robustness to imperfect CSI.https://spj.science.org/doi/10.34133/icomputing.0065 |
| spellingShingle | Yang Wang Zhen Gao Sheng Chen Chun Hu Dezhi Zheng Deep Learning–Based Channel Extrapolation and Multiuser Beamforming for RIS-aided Terahertz Massive MIMO Systems over Hybrid-Field Channels Intelligent Computing |
| title | Deep Learning–Based Channel Extrapolation and Multiuser Beamforming for RIS-aided Terahertz Massive MIMO Systems over Hybrid-Field Channels |
| title_full | Deep Learning–Based Channel Extrapolation and Multiuser Beamforming for RIS-aided Terahertz Massive MIMO Systems over Hybrid-Field Channels |
| title_fullStr | Deep Learning–Based Channel Extrapolation and Multiuser Beamforming for RIS-aided Terahertz Massive MIMO Systems over Hybrid-Field Channels |
| title_full_unstemmed | Deep Learning–Based Channel Extrapolation and Multiuser Beamforming for RIS-aided Terahertz Massive MIMO Systems over Hybrid-Field Channels |
| title_short | Deep Learning–Based Channel Extrapolation and Multiuser Beamforming for RIS-aided Terahertz Massive MIMO Systems over Hybrid-Field Channels |
| title_sort | deep learning based channel extrapolation and multiuser beamforming for ris aided terahertz massive mimo systems over hybrid field channels |
| url | https://spj.science.org/doi/10.34133/icomputing.0065 |
| work_keys_str_mv | AT yangwang deeplearningbasedchannelextrapolationandmultiuserbeamformingforrisaidedterahertzmassivemimosystemsoverhybridfieldchannels AT zhengao deeplearningbasedchannelextrapolationandmultiuserbeamformingforrisaidedterahertzmassivemimosystemsoverhybridfieldchannels AT shengchen deeplearningbasedchannelextrapolationandmultiuserbeamformingforrisaidedterahertzmassivemimosystemsoverhybridfieldchannels AT chunhu deeplearningbasedchannelextrapolationandmultiuserbeamformingforrisaidedterahertzmassivemimosystemsoverhybridfieldchannels AT dezhizheng deeplearningbasedchannelextrapolationandmultiuserbeamformingforrisaidedterahertzmassivemimosystemsoverhybridfieldchannels |