AI-based Olfactory Sensing for Space Application

This paper presents an AI-based olfactory sensing system designed for integration with robotic platforms and space rovers for planetary explorations. The system aims to enhance extraterrestrial operations through real-time analysis of atmospheric composition and detection of hazardous gases or substances. While traditional sensory systems such as vision, tactile sensors and laboratory-grade instruments have improved situational awareness, they face limited adaptability in unstructured and chemically dynamic extraterrestrial environments, that demand real time prediction. The proposed system utilizes advanced gas sensors combined with Machine Learning (ML) models which measure atmospheric concentrations of gases such as carbon monoxide (CO), nitrogen dioxide (NO2), ethanol (C2H6O) and a variety of volatile organic compounds (VOCs). By analyzing these primary gas signatures which are emitted by sources, the system enables real-time mapping of gas distribution and accurate classification of potentially targeted substances and gasses.A working prototype was developed using ESP32-based edge devices and Grove multichannel gas sensors, achieving over 82% classification accuracy with known target substances and gases. ML models, including Artificial Neural Networks (ANN) and Multinomial Naive Bayes (MNB) were applied to sensor data to ensure high precision and real-time response. The system is lightweight, low-power, and compact addressing the limitations of traditional sensors and lab equipment in terms of size, weight and energy consumption which is very critical in space exploration. Designed for seamless integration with space rover sensory systems, which may enables rovers to perform autonomous navigation and environmental sensing. It supports real-time atmospheric mapping, biosignature detection, environmental monitoring and also identification of various combustion fuel gases. These capabilities make the proposed system a valuable tool for future planetary missions where adaptive and intelligent sensing is essential.