Robot active olfaction search in turbulent flow and infotaxis search based on Rényi divergence

Searching for the source emitting chemical materials (pollutants, oil, toxic) in river/sea environments is particularly challenging given that the chemical transport is dominated by turbulent flow. Animals in nature can be very efficient at leveraging the olfaction to solve these problems e.g. foraging, mate-seeking, homing and host-seeking. As such, realizing active olfaction search on robotic platforms is plausible and has become a prominent research area in recent years. In this paper, we first review the various methods inspired by biology olfaction in the literatures and organize them into taxonomic classifications. The features and effectiveness of these methods are discussed and evaluated. Secondly, we investigate a novel infotaxis search strategy which measures the information gain utilizing Rényi divergence. Through selecting the parameter of Rényi divergence, this method pays more attention to regions of low-probability, which allows for the maximum discrimination between the priori and posterior probability of source likelihood distribution in the search progress. This feature makes the expected information reward over each action be obvious and the searchers avoid information direction losing in conventional infotaxis. Finally, the simulation results show infotaxis based on Rényi divergence reduces the search time and improves the search trajectory of the olfaction search.