A New Meta-Learning Framework for Estimating Atmospheric Turbulence and Phase Noise in Optical Satellite Internet-of-Things Systems

With the advantages of super-high transmission rate, anti-electromagnetic interference and good confidentiality, optical wireless communication (OWC) systems play an important role in satellite Internet of Things (IoT). In this paper, we propose a new meta-learning-based channel estimation scheme (Meta-CE) to address the challenges of atmospheric turbulence and phase distortion in satellite OWC links. A neural network-based channel estimator outputs channel parameters, rather than categories, and utilizes meta-learning to enhance its convergence speed and adaptability to new environments. The Meta-CE exhibits superior estimation accuracy, fast convergence, and generalization, compared to baseline schemes, and even outperforms the minimum mean square error (MMSE) channel estimation, especially with short pilot symbols, low signal-to-noise ratio (SNR) and severe turbulence. In a 4×4 multiple-input multiple-output (MIMO) scheme with 8-bit pilot symbols and 0 dB SNR, the mean square error of the Meta-CE is about 35% lower than that of the MMSE in a strong Gamma-Gamma turbulence channel.