Search papers, labs, and topics across Lattice.
This paper introduces a physics-informed learning approach to improve acoustic localization in challenging outdoor environments, addressing the limitations of traditional methods that struggle with multipath effects and near-field conditions. By refining a fast hyperbolic solver with a learned model that corrects implausible outputs, the method significantly reduces worst-case localization errors while maintaining median accuracy on field data. Additionally, it provides calibrated uncertainty estimates that enhance the reliability of spatial models in wildlife monitoring applications.
Traditional acoustic localization methods can fail dramatically in complex environments, but this new approach cuts worst-case errors while keeping median accuracy intact.
Recent advances in Passive Acoustic Monitoring (PAM) offer an opportunity to obtain ecological spatial point-process data at unprecedented scale. However, realizing this opportunity necessitates the development of accurate and scalable localization methods. In real-world outdoor soundscapes, however, the assumptions underlying classical localization methods such as hyperbolic and score-based localization are routinely violated by multipath dominance, near-field effects, and complex propagation. Under these conditions, classical localization methods become brittle, with extreme errors possible even in small detection arrays. Rather than statistically replacing the underlying physics, we propose a method to refine it and increase robustness outside of ideal operating conditions: a learned model operating on physics-informed acoustic features corrects a fast hyperbolic solver where it produces implausible solutions, substantially reducing catastrophic worst-case errors while matching its median accuracy on field data. We further provide calibrated, geometry-aware uncertainty estimates suitable for propagation into downstream spatial models. Evaluating on distributed microphone arrays in real and simulated outdoor environments, we demonstrate that the proposed method yields robust, uncertainty-aware localization, providing a step toward scalable automated wildlife monitoring in complex acoustic environments.
