Underwater Sound Speed Prediction by Mamba Neural Networks

The underwater sound speed profile (SSP), that describing the distribution of sound speed, is an important parameter of underwater communication positioning, navigation and timing system due to its great influence on the distribution of underwater sound field. Classically, SSPs can be obtained by measuring instruments or inversed based on sonar data, however, these methods need expensive equipment and are time-consuming. To avoid on-site data collection, we propose an underwater SSP prediction method by Mamba neural networks. We conducted experiments using the distribution of sound velocity in the Pacific Ocean as an example, and the results indicate that using historical SSP data, we can estimate the sound speed distribution more accurately compared to other popular time series prediction methods.