Joint Source-Environment Adaptation for Deep Learning-Based Underwater Acoustic Source Ranging

TL;DR

This work tackles the poor generalization of deep-learning–based underwater acoustic localization under environment mismatch by proposing a source-free domain adaptation framework. It introduces Joint Source-Environment Adaptation (JSEA), which fuses a neural prediction from SCM-based CNN features with a coarse energy-based cue derived from received signal strength, enabling robust localization without labeled target data. A complementary SHOT method provides partial adaptation via an information-maximization objective on a self-supervised subset. Empirical results on Bellhop data, reflecting a SWellEx-96 scenario with KAM11 noise, demonstrate that JSEA significantly improves MAE and PCL over SHOT and MFP under SSP and noise mismatches, highlighting the practical value of energy-aware probabilistic fusion in distributed underwater sensing.

Abstract

In this paper, we propose a method to adapt a pre-trained deep-learning-based model for underwater acoustic localization to a new environment. We use unsupervised domain adaptation to improve the generalization performance of the model, i.e., using an unsupervised loss, fine-tune the pre-trained network parameters without access to any labels of the target environment or any data used to pre-train the model. This method improves the pre-trained model prediction by coupling that with an almost independent estimation based on the received signal energy (that depends on the source). We show the effectiveness of this approach on Bellhop generated data in an environment similar to that of the SWellEx-96 experiment contaminated with real ocean noise from the KAM11 experiment.

Figures (6)

• Figure 1: An example of a predicted label in a test environment with a different sound speed profile than the training environment. Observe that the pre-trained model prediction (the label with maximum probability) is far from the ground truth, however, there is another peak in the output closer to the ground-truth.
• Figure 2: An example of a metric-inspired smoothed label $y_i$ corresponding to $d^q_i = 20$. Here, we have used the absolute error as the metric, hence, we have used a truncated exponential as the label.
• Figure 3: The CNN classifier includes a feature extraction part followed by a linear classifier that will be frozen during adaptation.
• Figure 4: Bellhop environment inspired by the SWellEx-96.
• Figure 5: Performance under different SNR values, with KAM11 noise and $\Delta c = 0.1$ m/s.