Mobile Robot Based Odor Source Localization In Outdoor Time-Variant Airflow Environments

In the world of living creatures, odor or pheromone is widely used for many activities, such as finding mates, searching for food, exchanging information, and evading predators. Inspired by the olfaction abilities of these creatures, in the early 1990s, researchers started trying to build mobile robots with onboard gas sensor(s) to locate the odor source. This research is often called odor source localization (OSL) or active olfaction.The research on the OSL is related to the fields such as fluid dynamics, sensing and information processing, bionics, computation intelligence, and mobile robot navigation and control. It is expected that the OSL will play more and more roles in such areas as fighting against terrorist attacks, finding toxic or harmful gas-leakage locations, checking for contrabands (e.g., heroin), searching for survivors in collapsed buildings or waters, and exploring mineral resources erupted in deep seas.This dissertation focuses on the OSL in outdoor time-variant airflow environments by using a single mobile robot. The main achievements can be concluded as follows:Firstly, in view of the problem that both the response time and recovery time of the commonly used metal-oxide-semiconductor gas sensors are relatively long, an adaptive-threshold algorithm is proposed to binarize the gas concentration. With the adaptive threshold, both the gas detection and non-detection events can arise in time, thus the robot can immediately and correctly judge the gas arrival and departure during the OSL.Secondly, a method for estimating an odor-patch path is put forward. This so-called odor-patch path is defined as the historical trajectory of the odor patch detected by the robot. The estimated odor-patch path can be applied to search for the odor source and finally to declare the odor source. The experimental results in outdoor airflow environments show that the estimated odor-patch path covers the odor source with a high probability.Thirdly, owing to the fact that the wind direction in outdoor environments always changes rapidly, a set of highly purposive search strategy for the odor source is proposed, which consists of a flow-direction-adaptive zigzagging-plume-finding behavior, an estimation-based plume-tracing behavior, and a biologically inspired plume-reacquiring behavior. The outdoor experiments demonstrate that with the proposed behaviors the robot can quickly find and effectively trace the plume, and also can reliably approach the odor source.Finally, a scheme for odor source declaration is proposed, which consists of a particle-filter-based method for the source-location estimation and a statistics-based rule for the source identification. The outdoor experiments demonstrate that the proposed estimation method is robust and the source location can be reliably estimated with a high accuracy, and whether the real source exists near the estimated location can be asserted with a high success rate using the proposed identification rule.