Joint Source-Environment Adaptation of Data-Driven Underwater Acoustic Source Ranging Based on Model Uncertainty

2025-03-30

AI Summary

Pretrained models for underwater acoustic source localization suffer from poor generalization to unseen oceanic environments and lack labeled data in the target domain. Method: This paper proposes an online joint source–environment adaptation method that requires neither target-domain labels nor access to original training data. Contribution/Results: Its core innovations include (i) the first use of implicit model uncertainty quantification for sample grouping and iterative pseudo-label refinement, and (ii) integration of signal energy–based independent estimation to enhance physical consistency. The uncertainty-driven adaptation mechanism enables zero-shot adaptation to unseen target environments. Evaluated on high-fidelity simulations and real sea-trial data, the method significantly improves acoustic source distance estimation accuracy—demonstrating robustness to complex, noisy, and distributionally shifted underwater conditions.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Intelligent Robots: State Estimation
• Search and Optimization: Learning to Search

Application Category

• User Modeling, Personalization and Recommendation: User modeling for targeted and personalized online advertising
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs

Abstract

Adapting pre-trained deep learning models to new and unknown environments is a difficult challenge in underwater acoustic localization. We show that although pre-trained models have performance that suffers from mismatch between the training and test data, they generally exhibit a higher ``implied uncertainty'' in environments where there is more mismatch. Leveraging this notion of implied uncertainty, we partition the test samples into more certain and less certain sets, and implement an estimation method using the certain samples to improve the labeling for uncertain samples, which helps to adapt the model. We use an efficient method to quantify model prediction uncertainty, and an innovative approach to adapt a pre-trained model to unseen underwater environments at test time. This eliminates the need for labeled data from the target environment or the original training data. This adaptation is enhanced by integrating an independent estimate based on the received signal energy. We validate the approach extensively using real experimental data, as well as synthetic data consisting of model-generated signals with real ocean noise. The results demonstrate significant improvements in model prediction accuracy, underscoring the potential of the method to enhance underwater acoustic localization in diverse, noisy, and unknown environments.

Problem

• Adapting pre-trained models to new underwater environments
• Reducing performance mismatch between training and test data
• Improving acoustic localization accuracy without target environment labels

Innovation

• Utilizes model uncertainty for sample partitioning
• Adapts pre-trained models without target labels
• Integrates signal energy for enhanced adaptation