Intelligent Flying Chemical Sensor Network for Gas-Leak Localization and Mapping

Abstract This article describes the development of a network of intelligent flying robotic sensors for quick and precise localization, estimation, and mapping of chemical-gas leaks. The key advances of the technology include leveraging decentralized Bayesian inference machine learning for source-term estimation, minimizing uncertainty through coordinated bio-inspired actions between robots, incorporating open-sector collision avoidance for safe autonomous navigation during search, and utilizing a kernel-based chemical distribution process to create chemical-concentration maps. Simulation and real-world outdoor-field tests with propane gas demonstrate the effectiveness of the flying-sensor network. This multirobot system can assist emergency responders in assessing and containing the spread of dangerous chemical leaks.