Robots with smell could be used to find sources of toxic gas, search for drugs, locate survivors under rubble or hidden mines. Robots have an advantage over sniffer dogs since they can enter dangerous areas and could potentially be deployed rapidly and at lower costs.
Most of the work on olfaction robots is done in controlled laboratory environments. Instead, Li et al. are looking at how robots can localize odor sources in outdoor environments with changing wind that can be turbulent and strong. To do this, they’ve developed a novel algorithm based on particle filters that tracks the location of the odor source over time.
Experiments were done using a two-wheeled robot equipped with a gas sensor, an anemometer (for wind measurements) and an electronic compass. The robot was placed in an outdoor 10 m × 10 m area and was asked to search for a humidifier containing liquid ethanol (odor source). To find the gas plume, the robot would perform spirals. As soon as gas was sensed, the robot followed a plume-tracing strategy to collect more information about the odor source. The videos below show the robot behavior (top) and the particle filter algorithm used (bottom).
Results show that the particle filter method is suitable for challenging outdoor odor source localization and that it outperforms Bayesian-inference-based methods.