Non-Terrestrial Networks (NTNs) in 6G are expected to integrate the Terrestrial Networks (TNs) coverage and provide connectivity to a multitude of User Terminals (UTs). In order to cope with the traffic demand, NTNs can employ cell- free MIMO with full frequency reuse schemes to maximize the spectral efficiency of the system. However, such technique can be strongly affected by channel aging, impacting their application to Low Earth Orbit (LEO)-based NTNs. Aiming at counteracting this effect, we present in this paper a lightweight Artificial Intelligence (AI) model for Channel State Information (CSI) prediction. To predict the propagation channel, the proposed algorithm learns its temporal statistics, e.g., the Line-of-Sight (LoS) shadowing correlation, from data. The model then applies the corrections to the feedback CSI, minimizing the difference with respect to the propagation channel encountered at transmission time. System-level analyses on a LEO-based CF-MIMO system report an improvement of the per-user capacity by up to 15% and a reduction of the outage probability when predicted CSI are used instead of aged channel coefficients.
Bruno De Filippo, Riccardo Campana, Alessandro Guidotti, Carla Amatetti, Alessandro Vanelli-Coralli (2024). Cell-Free MIMO in 6G NTN with AI-predicted CSI. Piscataway, NJ : Institute of Electrical and Electronics Engineers Inc. [10.1109/SPAWC60668.2024.10694298].
Cell-Free MIMO in 6G NTN with AI-predicted CSI
Bruno De Filippo
;Riccardo Campana;Alessandro Guidotti;Carla Amatetti;Alessandro Vanelli-Coralli
2024
Abstract
Non-Terrestrial Networks (NTNs) in 6G are expected to integrate the Terrestrial Networks (TNs) coverage and provide connectivity to a multitude of User Terminals (UTs). In order to cope with the traffic demand, NTNs can employ cell- free MIMO with full frequency reuse schemes to maximize the spectral efficiency of the system. However, such technique can be strongly affected by channel aging, impacting their application to Low Earth Orbit (LEO)-based NTNs. Aiming at counteracting this effect, we present in this paper a lightweight Artificial Intelligence (AI) model for Channel State Information (CSI) prediction. To predict the propagation channel, the proposed algorithm learns its temporal statistics, e.g., the Line-of-Sight (LoS) shadowing correlation, from data. The model then applies the corrections to the feedback CSI, minimizing the difference with respect to the propagation channel encountered at transmission time. System-level analyses on a LEO-based CF-MIMO system report an improvement of the per-user capacity by up to 15% and a reduction of the outage probability when predicted CSI are used instead of aged channel coefficients.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.