Experimental analysis of the impact of sensor response time on robotic fas source localization

Abstract

Robotic olfaction is a promising application of volatile-sensing technologies with many potential uses, such as gas leak detection or rescue missions. The effectiveness of any proposal to cope with this task is partially limited by the characteristics of the sensors employed. Most of the small, low-power sensing devices that are appropriate for mobile robotics suffer from problems like slow response time, long recovery, or low selectivity; and this makes the search process slower and less robust. In this work, we seek to provide a preliminary quantification of the effect of these issues in the results of a state-of-the-art gas source localization method. By using high-fidelity gas dispersion simulations, we compare the behaviour of the robotic agent when attaining measurements from a realistic sensor, with modelled response delay and inaccuracy, versus an idealized one that returns the ground-truth gas concentration value instantly. We find that, for slow-response sensors, there is a clear improvement of the results as the robot is allowed to gather measurements for a longer period of time, giving the sensor time to stabilize. A fast-responding sensor, on the other hand, does not benefit from very long measurement steps, needing only a small time window to account for instantaneous instabilities in the spatial dispersion of the gas.