A 3D Spiral-Zigzag Gas Leakage Source Search Algorithm Based on Olfaction and Vision

Accidents such as fires and chemical leakage, which occur from time to time, seriously endanger public safety. The use of robot to discover and track the source of gas/chemical plume can replace traditional manual search, greatly reducing the risk and cost of search and improving efficiency. Most of the existing gas leakage source search methods are applied in two-dimensional environments, and for the scenarios where the plume is visible, the efficiency of searching and the success rate of source localization only relying on the olfaction is usually low. To this end, this paper proposes a plume finding and tracing method based on olfaction and vision with a drone in 3D outdoor environments. In the plume finding phase, olfactory information is combined with visual object detection in the 3D Spiral algorithm to improve plume finding efficiency. During the plume tracing phase, the drone performs a 3D Zigzag motion to rapidly approach the source of the gas leakage. If the plume is lost during tracing, the 3D Spiral algorithm, which combines olfactory and visual information, is used again in local area with a view to find the plume again. Experimental and simulation results show that the efficiency of plume finding can be significantly improved by using visual object detection for the plume, and the proposed plume finding and tracing method combining olfactory and visual information has high search efficiency and success rate.