A Perception and Mapping Approach for Plume Detection in Payload Directed Flight

摘要

Payload Directed Flight (PDF) is a research task under NASA Aeronautics Research Mission Directorate, Fundamental Aeronautics Program, Subsonic Fixed Wing Project. In recent years, the capabilities, cardinality, and utilization of onboard sensor and payload suites have been greatly increasing throughout the aviation industry in various flight applications. Through intelligent coupling between the autopilot, onboard and payload sensors, next generation aircraft can take advantage of greater data availability to better achieve their mission objectives. Whether the objectives are mapping, tracking, or surveillance, the ability to close the loop around payload and non-traditional sensors using intelligent variably-autonomous algorithms enables increased efficiency and performance as well as exciting next-generation capabilities to vehicle platforms. Mapping and surveillance of dynamic and amorphous phenomena remains a difficult problem in the perception area of PDF research. As an example application, this research examines an algorithm for wildfire plume mapping by extending recent work in computer vision and robotic mapping. In a realistic simulation of a plume-tracking mission, results demonstrate that the vehicle was able to successfully map and track a dynamic plume in real-time to approximate visual accuracy.