Semantic Communication in Underwater Communication: Advantage, Problem and Solution — A survey

Underwater acoustic channel has large noise, serious fading and narrow available bandwidth, resulting in poor reliability and low efficiency of underwater communication. Semantic communication adopts machine learning technology, which will provide more reliable and efficient communication than traditional communication in the face of low signal-to-noise ratio ( SNR ) and narrow bandwidth channel environment. Therefore, it is expected to be applied to underwater acoustic communication in the future. In this paper, a semantic communication model based on convolutional autoencoder is used for simulation analysis, and an untrained layer representing underwater acoustic communication channel is set in the middle. The results show that semantic communication has better performance in underwater acoustic channel. Due to the time-varying characteristics of underwater acoustic channels, in order to ensure the quality of service ( Qos ), it is necessary to update the semantic model in time, which is a difficult challenge in the face of complex and special underwater environment. For this, we propose solutions to provide ideas for future research in this field.