Compact and embedded electronic nose for volatile and non-volatile odor classification for robot applications

Abstract

Electronic noses are studied and developed since many years, aiming today to enhance the sensitivity floor, the response time, or characterize new chemical processes. Nowadays, the most performant apparatus are cumbersome, expensive, and not fully dedicated to mobile systems. Most of the researches related to embedded noses on moving applications aim to develop mapping or source detection of toxic gases, enhancing the geometry of the nose and taking into account the air flow perturbations. In this study, we aimed to develop a compact electronic nose dedicated to identify volatile and nonvolatile odors. The aim is at midterm to embed it on a humanoid robot. The purpose is to achieve a compact and embedded electronic nose to provide an added function for robots evolving in their daily environment and able to learn more odors. The final goal is mainly to use it in human-like behavior, mostly to sense food with several conditioning possibilities like volatile or nonvolatile odors. The developed E-Nose has four MOS gas sensors, embedded electronics, and software based on k-nearest neighbors' classification algorithm. Experimental results show success rates up to 98% to differentiate between four fruits juices, namely apple, orange, pineapple, and grenade and to identify rotten eggs from good eggs, grenade perfume, and butane gas, in less than 60 s. An extensible interface enables the E-Nose to learn more odors.