Bees track down buried explosives, dogs the hideout of escaped convicts - following a scent seems like child's play. But sending sniffing robots on a search is by no means trivial. The math shows where to go.
Maybe still looking for the station bakery or for "what smells so burnt there" - otherwise people rarely use their nose as a guide. In the animal world, on the other hand, the sense of smell is an indispensable tool for finding food, escaping enemies or tracking down conspecifics that are willing to mate: salamanders follow scent trails in search of food, bees leave scent marks on flowers, aphids jump off plants when a ladybug smelling suspiciously is approaching, female butterflies lure their admirers with the sparsest of scent clouds.
If you only rely on seeing and hearing, you can miss a lot. Objects can be obscured or well camouflaged. Noises can be confused, drowned out or distorted by echoes. When looking for a specific object, smelling, i.e. detecting certain chemical substances, has an advantage - provided the object you are looking for emits an odour: scents are very specific because the range of chemical components is almost unlimited.
One clear disadvantage of smell, however, is the lack of directional information: it is difficult for us to determine where an odor is coming from. If you want to follow a scent trail, you usually fail because it is divided into separate clouds of scent and spreads randomly in unpredictable currents. However, many animals are able to reliably follow odor sources solely on the basis of the scent concentration and spatial distribution of the clouds as well as the wind direction.
Test objects that are particularly popular with researchers include the two butterfly species silk moth (Bombyx mori) and tobacco hawkmoth (Manduca sexta). When males notice the enticing pheromone clouds of the females, they begin a zigzag flight against the wind. If they don't find a new cloud of scent for a long time, they circle in wide circles or fly sideways to the wind without moving forward. Rediscover the fragrance, then switch from this "casting" strategy back to the "zig zag" mode. Crucial to success is the balance between the two strategies "exploration" and "exploitation": exploring the environment on the one hand to gain information about the location of the source and on the other hand moving in order to get in the suspected direction of the source.
A French-American research team led by Massimo Vergassola from the Pasteur Institute in Paris regards the search as an information-theoretical problem and has now presented an algorithm that describes the - mathematically speaking - optimal procedure. A map is calculated from the distribution of the odor clouds found so far, which indicates the probability of the source of the odor being located at a specific location in the room. This spatial probability map has "entropy" - a measure of the seeker's uncertainty. Initially, little information is available about the location of the source of the odor, so the probability map remains fuzzy and the entropy high.
In order to find the source as quickly as possible, the entropy must be reduced as much as possible with each search step. At the same time, the probability map becomes sharper; the information regarding the location of the source of the scent is constantly growing. Based on the term "chemotaxis" - the movement of organisms controlled by substance concentrations - the scientists call this method "infotaxis". The algorithm automatically ensures the optimal mix of "exploration" and "exploitation". In theoretical simulations, the "infotaxis" was much faster than if the two search strategies alternated.
The theoreticians have done their job, but there is still a long way to go before the perfect robot nose. In terms of sensitivity, versatility and speed, chemical sensors lag far behind animals. In addition, additional information such as wind direction and strength and visual information must also be processed and reliably combined. Added to this is the limited navigational ability of today's robots.
A research group at the SPECS laboratory at the Pompeu Fabra University in Barcelona has already got many of these problems under control. They are currently working on a small unmanned airship called "AMOTH" (Artificial Moth) that will be used to search for mines or toxic chemicals. Their construction is based on the behavior and neurophysiology of butterflies. While the search for an odor source in the laboratory works very reliably, successful outdoor use is still pending.
There is still a lot of research to be done before smelling search robots reach the maturity required for commercial applications. Until then, we'll have to rely on animal noses.