Deep Probabilistic Indoor Gas Source Localization via Physical Dependency-Guided Sequential Inference

2026-08-17

AI Summary

This study addresses the challenges of gas source localization under complex indoor airflow and the computational burden of high-fidelity models by proposing a physics-informed deep probabilistic framework. The method embeds gas transport physics into sequential conditional reasoning, integrating wind and concentration fields to guide posterior source estimation while employing an active search strategy for precise online localization via mobile robots under sparse, noisy observations. Experimental results demonstrate that this framework significantly outperforms mainstream baselines, achieving both high simulation fidelity and computational efficiency. Furthermore, real-world validation on embedded GPUs confirms its capability for online operation, effectively reconciling the trade-off between physical model accuracy and real-time inference requirements in dynamic indoor environments.

Abstract

Reliable gas source localization (GSL) is critical to safety in industrial and urban environments, yet remains challenging indoors because walls and obstacles interact with airflow to create complex gas dispersion. High-fidelity models such as computational fluid dynamics and filament models can capture these effects, but their computational cost limits online use. We propose a deep probabilistic framework that infers the source posterior from sparse and noisy measurements collected by a mobile robot. Unlike end-to-end models that directly infer source estimates from measurements, the proposed method incorporates physical dependencies of indoor gas transport, where wind and source location govern the concentration field. These dependencies are embedded through sequential conditional inference, in which inferred wind and concentration fields guide source posterior estimation. This structure improves localization under sparse and noisy observations. Evaluations show that the proposed method outperforms representative GSL baselines and enables accurate and efficient active GSL in simulations. Real-robot experiments demonstrate the feasibility of online operation on an embedded GPU.

Problem

• Gas Source Localization
• Indoor Environment
• Sparse and Noisy Measurements
• Online Operation

Innovation

• Deep Probabilistic Framework
• Physical Dependency-Guided Inference
• Sequential Conditional Inference
• Gas Source Localization
• Online Active GSL