Title: Diffusion Graph Neural Networks for Robustness in Olfaction Sensors and Datasets Show Chinese title

Title: 用于嗅觉传感器和数据集鲁棒性的扩散图神经网络

Abstract: Robotic odour source localization (OSL) is a critical capability for autonomous systems operating in complex environments. However, current OSL methods often suffer from ambiguities, particularly when robots misattribute odours to incorrect objects due to limitations in olfactory datasets and sensor resolutions. To address this challenge, we introduce a novel machine learning method using diffusion-based molecular generation to enhance odour localization accuracy that can be used by itself or with automated olfactory dataset construction pipelines. This generative process of our diffusion model expands the chemical space beyond the limitations of both current olfactory datasets and training methods, enabling the identification of potential odourant molecules not previously documented. The generated molecules can then be more accurately validated using advanced olfactory sensors, enabling them to detect more compounds and inform better hardware design. By integrating visual analysis, language processing, and molecular generation, our framework enhances the ability of olfaction-vision models on robots to accurately associate odours with their correct sources, thereby improving navigation and decision-making through better sensor selection for a target compound in critical applications such as explosives detection, narcotics screening, and search and rescue. Our methodology represents a foundational advancement in the field of artificial olfaction, offering a scalable solution to challenges posed by limited olfactory data and sensor ambiguities. Code and data are made available to the community at the following URL: https://github.com/KordelFranceTech/OlfactionVisionLanguage-Dataset. Show Chinese abstract

Abstract: 机器人气味源定位（OSL）是自主系统在复杂环境中运行的关键能力。然而，当前的OSL方法常常面临歧义问题，特别是在机器人由于嗅觉数据集和传感器分辨率的限制而将气味错误地归因于不正确的物体时。为了解决这一挑战，我们引入了一种新的机器学习方法，使用基于扩散的分子生成来提高气味定位的准确性，该方法可以单独使用或与自动嗅觉数据集构建流水线结合使用。我们的扩散模型的生成过程扩展了化学空间，超越了当前嗅觉数据集和训练方法的限制，使得能够识别之前未记录的潜在气味分子。然后可以使用先进的嗅觉传感器更准确地验证生成的分子，使其能够检测更多化合物，并指导更好的硬件设计。通过整合视觉分析、语言处理和分子生成，我们的框架增强了机器人嗅觉-视觉模型准确关联气味与其正确来源的能力，从而通过为关键应用中的目标化合物选择更好的传感器来改善导航和决策，例如爆炸物检测、毒品筛查和搜索救援。我们的方法代表了人工嗅觉领域的一个基础性进展，为有限的嗅觉数据和传感器歧义带来的挑战提供了可扩展的解决方案。代码和数据可在以下网址提供给社区：https://github.com/KordelFranceTech/OlfactionVisionLanguage-Dataset.