Sound speed profile inversion based on distributed networked underwater sensors system and graph attention networks

Area of Science:

• Oceanography
• Acoustics
• Sensor Networks

Background:

• Variations in sound speed profiles (SSPs) critically impact underwater acoustic systems, causing sound ray bending that reduces communication and localization accuracy.
• Ocean acoustic tomography (OAT) offers a method for SSP estimation, but traditional techniques using linear arrays are difficult to scale for distributed networks.

Purpose of the Study:

• To develop an effective and scalable SSP inversion scheme for distributed networked underwater sensors (DNUS) systems.
• To improve the accuracy of underwater acoustic communication and localization by accurately estimating SSPs.

Main Methods:

• A novel SSP inversion scheme is proposed, integrating multimodal inputs including time difference of arrival, angle of arrival, node positions, and boundary sound speed into a sensing matrix.
• A graph attention network model is employed to learn the complex mapping between these input features and the SSP.

Main Results:

• The proposed method demonstrates effectiveness and accuracy in estimating SSPs within DNUS systems.
• Numerical simulations and a shallow-water validation experiment confirm the performance of the developed inversion scheme.

Conclusions:

• The developed SSP inversion scheme offers a robust solution for distributed underwater sensor networks.
• This advancement has significant implications for enhancing the performance of underwater acoustic technologies.