A softgrowing robotic system for odor detection and classification

Basit Öğe Kaydı
dc.contributor.authorOyejide, Ayodele James
dc.contributor.authorAstar, Ahmet
dc.contributor.authorKaya, Gulnur
dc.contributor.authorYaqub, Ustaz Abdulfattah
dc.contributor.authorBaran, Eray A.
dc.contributor.authorStroppa, Fabio
dc.date.accessioned2026-04-04T18:55:37Z
dc.date.available2026-04-04T18:55:37Z
dc.date.issued2026
dc.departmentİstanbul Bilgi Üniversitesi
dc.description.abstractOdor classification is essential in environmental monitoring, gas leak detection, and industrial safety. Although conventional mobile robotic platforms equipped with electronic noses offer advanced gas-sensing capabilities, their performance in confined or cluttered environments is often constrained by rigid structures and limited maneuverability. In this work, we present an olfactory softgrowing robot (oSGR) that integrates bio-inspired, growth-based locomotion with machine-learning (ML)-driven odor classification. Our system comprises a pressurized base enabling contact-free eversion and a custom motorized tip mount housing a multi-sensor array of four metal oxide TGS sensors (2600, 2602, 2611, and 2620) coupled with a passive aspirator for volatile organic compound (VOC) sampling. We provide detailed modeling, design, and structural characterization of the tip mount under multiple actuation configurations, and demonstrate the robot's olfactory capability through experiments involving four VOCs - ethanol, methane, gin, and acetone. We evaluated two experimental modes: (i) in-transit and static sampling at fixed distances ( $20$ , $40$ , and $80$ cm from the source), and (ii) continuous sampling during transit at speeds of $5$ cm/s and $10$ cm/s. The collected olfactory dataset was used to train twelve widely employed supervised ML classifiers in gas sensing, including k-Nearest Neighbors (kNN), Random Forest, and Linear Discriminant Analysis. The kNN classifier achieved the highest accuracy (99.88%), demonstrating strong robustness for the olfactory data. Our results highlight the potential of SGRs for contact-free, continuous, in-motion chemical sensing. This unique data acquisition approach reduces detection latency and energy consumption typically associated with conventional stop-and-sense strategies.
dc.description.sponsorshipTUBITAK [121C145]
dc.description.sponsorshipThis work is funded by TUB & Idot;TAK within the scope of the 2,232-B International Fellowship for Early Stage Researchers Program number 121C145.
dc.identifier.doi10.1017/S0263574726103257
dc.identifier.doi10.1017/S0263574726103257
dc.identifier.issn0263-5747
dc.identifier.issn1469-8668
dc.identifier.scopus2-s2.0-105032675215
dc.identifier.scopusqualityQ1
dc.identifier.urihttps://doi.org/10.1017/S0263574726103257
dc.identifier.urihttps://hdl.handle.net/11411/10490
dc.identifier.wosWOS:001709491200001
dc.identifier.wosqualityQ3
dc.indekslendigikaynakWeb of Science
dc.indekslendigikaynakScopus
dc.language.isoen
dc.publisherCambridge Univ Press
dc.relation.ispartofRobotica
dc.relation.publicationcategoryMakale - Uluslararası Hakemli Dergi - Kurum Öğretim Elemanı
dc.rightsinfo:eu-repo/semantics/openAccess
dc.snmzKA_WoS_20260402
dc.snmzKA_Scopus_20260402
dc.subjectBiologically Inspired Robots
dc.subjectE-Nose
dc.subjectMachine Learning
dc.subjectSoft Growing Robots
dc.subjectOlfactory Robotics
dc.titleA softgrowing robotic system for odor detection and classification
dc.typeArticle

Dosyalar

Koleksiyon

Web of Science Indexed Publications
Scopus Indexed Publications