Mismatch-Robust Underwater Acoustic Localization Using A Differentiable Modular Forward Model

2025-03-30

AI Summary

To address degraded robustness in underwater acoustic source localization caused by environmental model mismatch, this paper proposes an inference-time joint optimization method based on a differentiable modular forward model: during testing, it simultaneously optimizes both network weights and source location, enabling physics-driven end-to-end multipath modeling. This work is the first to integrate online meta-optimization (i.e., inference-time weight adaptation) with physics-informed modular modeling, enabling unsupervised learning of multipath propagation distances without explicit path-label supervision. Theoretical analysis establishes sufficient conditions for convergence and robustness. Experiments demonstrate substantial improvements in localization accuracy under train-test environmental mismatch. Moreover, the modular architecture facilitates implicit representation of complex multipath structures and enables cross-environment generalization.

Technology Category

• Search and Optimization: Learning to Search
• Machine Learning: Optimization
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

In this paper, we study the underwater acoustic localization in the presence of environmental mismatch. Especially, we exploit a pre-trained neural network for the acoustic wave propagation in a gradient-based optimization framework to estimate the source location. To alleviate the effect of mismatch between the training data and the test data, we simultaneously optimize over the network weights at the inference time, and provide conditions under which this method is effective. Moreover, we introduce a physics-inspired modularity in the forward model that enables us to learn the path lengths of the multipath structure in an end-to-end training manner without access to the specific path labels. We investigate the validity of the assumptions in a simple yet illustrative environment model.

Problem

• Underwater acoustic localization with environmental mismatch
• Optimize neural network weights during inference
• Learn multipath structure without path labels

Innovation

• Gradient-based optimization with neural network
• Simultaneous optimization of network weights
• Physics-inspired modular forward model