Effect of the agent size on the performance of an infotactic and a hybrid olfactory search based on the burstiness of odor pulses

Abstract

Being able to find odor sources is a desirable skill in autonomous robots with potential applications searching for hazardous substances in airports, disaster areas, industrial facilities, etc. Owing to the difficulty of olfactory searches, previous studies have taken a bio-mimetic approach. Other studies have developed probabilistic strategies; but few have explored the combination of both. In this study, we analyze a probabilistic strategy known as infotaxis. Here, we find a case where if the size of the searching agent increases, relative to the width of the environment, the performance of infotaxis decreases. To compensate for this disadvantage, we propose a hybrid algorithm. The hybrid algorithm starts an olfactory search with infotaxis and switches to a silk moth-mimetic algorithm when the agent is nearby the source and the signal of odor pulses is periodic. We evaluated the performance of the hybrid algorithm by simulations under video recordings of real odor plumes. As a result, we found that the hybrid algorithm significantly improved the success rate at the agent size that decreased the performance of infotaxis.