Research · Perception · Machine olfaction
Machine vision and hearing are solved; smell is the last major sense still missing from AI. An electronic nose copies biology's trick — not one detector per odour, but an array of imperfect, overlapping sensors read as a pattern by AI. What differs between approaches is only the physics of the single sensor: how a captured molecule becomes an electrical signal.
Four ways to turn a molecule into a signal, from today's workhorse to tomorrow's frontier. All run as arrays — the intelligence is in reading the whole pattern, not any one sensor. These are the approaches that matter for a deployable nose; it is not an exhaustive list.
Maturity — deployable today · Opportunity — untapped headroom. Ordinal 1–5, our synthesis of the cited reviews — not a measured index.
Every e-nose is the same pipeline. The six approaches are interchangeable front-ends that plug into the same array-plus-AI backbone — which is why the model and its data, not the sensor, carry the value.
Combinatorial coding. Your nose has ~400 receptor types yet distinguishes millions of smells — no receptor is specific to one odour; the brain reads the pattern across all of them. An e-nose does the same with a dozen cross-reactive sensors and a classifier.