Experimental Results of Underwater Sound Speed Profile Inversion by Few-shot Multi-task Learning

TL;DR

This work tackles SSP inversion under data-scarce conditions by introducing a multi-task learning framework that uses partial parameter sharing to extract common SSP features and accelerate learning. The method includes offline meta-training across multiple SSP clusters, dynamic task-specific learning rates based on spatio-temporal distance, and a transfer of shared representations to a task-specific learner. To address depth coverage gaps from XCTD data, the paper also proposes EOF-based SSP extension (EOF–PSSP–ME) for full-depth reconstruction. Validation from a deep-ocean experiment in the South China Sea shows that MT L outperforms state-of-the-art methods (MFP, CS, FNN) in accuracy while retaining the real-time, single-forward-pass inversion capability, highlighting its practical potential for data-scarce regions and operational underwater systems. $SSP$ inversion under sparse data conditions is thus improved, enabling more reliable underwater navigation and sensing.

Abstract

Underwater Sound Speed Profile (SSP) distribution has great influence on the propagation mode of acoustic signal, thus the fast and accurate estimation of SSP is of great importance in building underwater observation systems. The state-of-the-art SSP inversion methods include frameworks of matched field processing (MFP), compressive sensing (CS), and feedforeward neural networks (FNN), among which the FNN shows better real-time performance while maintain the same level of accuracy. However, the training of FNN needs quite a lot historical SSP samples, which is diffcult to be satisfied in many ocean areas. This situation is called few-shot learning. To tackle this issue, we propose a multi-task learning (MTL) model with partial parameter sharing among different traning tasks. By MTL, common features could be extracted, thus accelerating the learning process on given tasks, and reducing the demand for reference samples, so as to enhance the generalization ability in few-shot learning. To verify the feasibility and effectiveness of MTL, a deep-ocean experiment was held in April 2023 at the South China Sea. Results shows that MTL outperforms the state-of-the-art methods in terms of accuracy for SSP inversion, while inherits the real-time advantage of FNN during the inversion stage.

Figures (9)

• Figure 1: MTL SSP inversion model.
• Figure 2: Structure of neural networks in MTL.
• Figure 3: System composition of ocean experiments.
• Figure 4: Relative azimuth of GPS and USBL in horizontal direction.
• Figure 5: Space distribution of sampled SSPs.