Controlling a gas/odor plume-tracking robot based on transient responses of gas sensors

Abstract

Gas sensors provide an artificial sense of smell for a mobile robot to track an airborne gas/odor plume and locate its source. However, a slow response of gas sensors has been the major factor limiting the development of plume-tracking robots. This paper describes a new control algorithm that breaks the limitation. The basic idea is to detect onsets of gas sensor response and starts of recovery by monitoring the relative change in each sensor output. Fast plume tracking is accomplished by making the robot take appropriate actions immediately when the sensor outputs start changing from one state to another. Growing sensor outputs evoke an increase in the robot speed for further acceleration of plume tracking, whereas insufficient sensor outputs slow down the robot to avoid degrading the search success rate. In contrast to the previous algorithm, based on the absolute sensor output levels, the detection of output change also leads to reliable plume detection, since it is insusceptible to drift in the gas sensor outputs. Experimental results have shown that the robot can track down a gas source within the distance of 2 m in 32 s, even though semiconductor gas sensors with a long recovery time (>60 s) are used.