TruffleBot: Low-Cost Multi-Parametric Machine Olfaction

This paper presents a low-cost and flexible platform for bio-inspired machine olfaction, which aims to extend traditional electronic nose approaches by adding fluid-mechanical and spatiotemporal dimensions. The
TruffleBot
contains an array of chemical, pressure, and temperature sensors in a small embedded platform. By “sniffing” vapors in a temporally-modulated sequence through four different air paths across eight sensor locations, we introduce spatial and temporal information that significantly enhances classification of odors. Using only chemical time series, we demonstrate 91% cross-validated classification accuracy for nine odors. With the addition of pressure and temperature time series, TruffieBot's classification accuracy can reach 95-98%.