Collaborative Research: NRI: Integration of Autonomous UAS in Wildland Fire Management

点击翻译按钮获取中文摘要

This research project, in cooperation with the Ohio Department of Natural Resources (Division of Forestry), focuses on autonomous unmanned aerial systems (UAS) for operations in hazardous environments to perform wildfire monitoring during prescribed burns for fire prevention and mitigation. Climate change in the US has exacerbated wildfires and intensified the Department of Natural Resources activities in response. Experts from the areas of forest management and ecology, uncertainty quantification, sensor fusion and data-driven modeling and control collaborate to deploy autonomous aerial robotic systems in unstructured, uncertain, and hazardous fire environments. The research from these collaborations aids wildland-urban planning, preparing for and sustainment of a safe wildland fire response; in particular, this research contributes to understanding how topographic, atmospheric and forest fuel factors in temperate hardwood forests influence fire intensity and rate of spread. This project invites and encourages students to participate in robotics research. Through its outreach activities, the project also informs the general public of the value of robotics research for addressing societal challenges.Theoretical, computational, and experimental methods and materials developed in this work enhance situational awareness and enables autonomous risk-aware decision-making in unstructured and uncertain hazardous environments. UAS path planning will formulate and solve novel resource chance-constrained optimization problems. UAS will bypass computational heavy lifting to generate in-time micro-level local conditions by enabling physics-informed learning through Koopman operator theory. New sensor belief functions will be designed that accurately reflect sensing ignorance contained in hypotheses related to the fire environment. Evidential information fusion will effectively handle sensor epistemic uncertainty and allow reliable integration in an environment where not all data is trustworthy. Data-driven control will enable efficient and reliable operation of autonomous vehicles with uncertain dynamics in real time by using available knowledge of applied inputs and observed outputs, to learn the unknown inputs even without prior training data or persistent excitation. Real-time estimates of disturbance forces and torques acting on an UAS obtained by the disturbance observer will provide information on the turbulence and air flow around a wildland fire region.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.