Odor plumes are chemophysical fields where molecular identity is coupled to transport phenomena including diffusion and turbulence. In animals, decoding this joint information underlies navigation and social behavior. However, artificial systems treat identity and location as independent problems, solved by distinct modalities. Here we present AROMA (Artificial Receptor-Olfaction Mimetic Array), a stereo olfaction strategy in which receptor-mimetic sensors capture plume dynamics, enabling concurrent decoding of mixture composition and 3D source location from analyte-dependent kinetic patterns. AROMA is built on mixed-ligand gold nanoparticles whose phase-separated monolayers and non-additive interfacial energetics emulate the promiscuous selectivity of olfactory GPCRs. Deployed in an antenna-like, spatially separated configuration, these sensors convert evolving concentration fields into disparities in onset, rise, and amplitude that encode geometric information. Trained on 50 plume trials of six-odorant mixtures sampled by a 30-channel array in an 18 cm cubic chamber, a multi-task Transformer achieves 86.7% mixture identification accuracy and a 3D localization error of 2.84 ± 0.87 cm. We demonstrate room-scale source tracking with a mobile robot indoors under natural airflow. By translating plume dynamics into a unified latent representation, AROMA advances artificial olfaction from static molecular discrimination toward spatial intelligence, offering a general framework for environmental monitoring and autonomous navigation.
Yu et al. (Wed,) studied this question.