Robust Moth-Inspired Algorithm for Odor Source Localization Using Multimodal Information

Abstract

Odor-source localization, by which one finds the source of an odor by detecting the odor itself, is an important ability to possess in order to search for leaking gases, explosives, and disaster survivors. Although many animals possess this ability, research on implementing olfaction in robotics is still developing. We developed a novel algorithm that enables a robot to localize an odor source indoors and outdoors by taking inspiration from the adult male silk moth, which we used as the target organism. We measured the female-localization behavior of the silk moth by using a virtual reality (VR) system to obtain the relationship between multiple sensory stimuli and behavior during the localization behavior. The results showed that there were two types of search active and inactive depending on the direction of odor and wind detection. In an active search, the silk moth moved faster as the odor-detection frequency increased, whereas in the inactive search, they always moved slower under all odor-detection frequencies. This phenomenon was constructed as a robust moth-inspired (RMI) algorithm and implemented on a ground-running robot. Experiments on odor-source localization in three environments with different degrees of environmental complexity showed that the RMI algorithm has the best localization performance among conventional moth-inspired algorithms. Analysis of the trajectories showed that the robot could move smoothly through the odor plume even when the environment became more complex. This indicates that switching and modulating behavior based on the direction of odor and wind detection contributes to the adaptability and robustness of odor-source localization.