TY - JOUR
T1 - Channel-Adaptive Autoencoder Architecture for Terahertz Aerospace Communications
AU - Huang, Yicheng
AU - Mao, Tianqi
AU - He, Dongxuan
AU - Wang, Qi
AU - Han, Chong
AU - Wang, Zhaocheng
N1 - Publisher Copyright:
© 2025 IEEE.
PY - 2026
Y1 - 2026
N2 - The terahertz (THz) aerospace communication has garnered increasing attention due to its potential to meet the ultra-broadband transmission demand of space-air backbone networks. However, several practical issues have to be considered, which include the complex channel characteristics, antenna misalignment, hardware imperfections, etc. To tackle these issues, a dynamic autoencoder system based on channel prediction is proposed to enhance its robustness against complicated dynamic channel fading and distortions in aerospace THz communication systems. Specially, a new architecture with a set of autoencoders and fitting networks is proposed for resilient transmission against the strong Doppler shifts and hardware imperfections. Besides, a single point channel prediction model is utilized to forecast the variations of the channel and then subsequently select the suitable autoencoder. Numerical results demonstrate the superiority of the proposed dynamic autoencoder architecture under an intricate dynamic THz channel.
AB - The terahertz (THz) aerospace communication has garnered increasing attention due to its potential to meet the ultra-broadband transmission demand of space-air backbone networks. However, several practical issues have to be considered, which include the complex channel characteristics, antenna misalignment, hardware imperfections, etc. To tackle these issues, a dynamic autoencoder system based on channel prediction is proposed to enhance its robustness against complicated dynamic channel fading and distortions in aerospace THz communication systems. Specially, a new architecture with a set of autoencoders and fitting networks is proposed for resilient transmission against the strong Doppler shifts and hardware imperfections. Besides, a single point channel prediction model is utilized to forecast the variations of the channel and then subsequently select the suitable autoencoder. Numerical results demonstrate the superiority of the proposed dynamic autoencoder architecture under an intricate dynamic THz channel.
KW - Aerospace communication
KW - antenna misalignment
KW - autoencoder
KW - high mobility
KW - terahertz communication
UR - https://www.scopus.com/pages/publications/105013986192
U2 - 10.1109/TVT.2025.3601728
DO - 10.1109/TVT.2025.3601728
M3 - Article
AN - SCOPUS:105013986192
SN - 0018-9545
VL - 75
SP - 3288
EP - 3293
JO - IEEE Transactions on Vehicular Technology
JF - IEEE Transactions on Vehicular Technology
IS - 2
ER -