Researchers at the Pasteur Institute in Paris have developed an algorithm that allows robots to find the source of a faint scent even with air turbulence. Like a moth would do.

Massimo Vergassola and his colleagues tested their algorithm in a virtual environment and they discovered that not only the robot (virtual, also!) succeeds to find the source but overall, it moves in complex back and forth sweeping motions, s-curves, and spirals that closely look like the way of a moth.

The algorithm uses information received from the scent itself but when no scent is detected, it mixes to behaviours: going directly toward the point where it guesses the scent is coming from and wandering around collecting information.

Vergassola says that the algorithm could be implemented in an real sniffer robot but also in any other application that involves searching without much information. Example: detecting the best paths for data to be sent through a network.


Source: NewScientistTech, last week.