Hankel-FNO: Fast underwater acoustic charting via physics-encoded Fourier neural operator

Area of Science:

• Ocean acoustics
• Computational mathematics
• Artificial intelligence

Background:

• Accurate underwater acoustic charting is vital for oceanographic applications like sensor placement and autonomous vehicle navigation.
• Traditional acoustic charting methods are computationally intensive and not suitable for real-time or large-scale use.
• Existing deep learning models face limitations in resolution and reliance on explicit physical models, hindering their generalizability.

Purpose of the Study:

• To develop an efficient and accurate computational model for underwater acoustic charting.
• To overcome the limitations of conventional solvers and existing deep learning approaches.
• To enhance the scalability and applicability of acoustic charting in diverse marine environments.

Main Methods:

• Proposed Hankel-Fourier Neural Operator (Hankel-FNO), a deep learning model based on Fourier Neural Operators.
• Integrated physical knowledge of sound propagation and bathymetry into the model architecture.
• Validated the model's performance against traditional numerical solvers and data-driven alternatives.

Main Results:

• Hankel-FNO achieved superior accuracy and computational speed compared to conventional methods.
• The model demonstrated higher accuracy than existing data-driven alternatives, particularly for long-range acoustic predictions.
• Experimental results confirmed the model's adaptability to various environments and sound source configurations with minimal fine-tuning.

Conclusions:

• Hankel-FNO offers a significant advancement in underwater acoustic charting, balancing speed and accuracy.
• The model's physics-informed approach enhances its generalization capabilities across different oceanographic scenarios.
• This method provides a scalable and efficient solution for critical underwater mapping and navigation tasks.