“Autonomous Flying Fire Blanket”: New Adaptive And Learning Architectures For Multi-UAV Cooperative Formation With Firefighting Applications

点击翻译按钮获取中文摘要

Fires cause significant damage in the United States every year. Unmanned aerial vehicles have been increasingly used in fire fighting and prevention tasks. However, current firefighting aerial vehicles suffer from several limitations and constraints, including complex designs of the platforms, limited tank capacity and flow rate for the suppressant fluid, and high costs. This project aims to develop a new “autonomous flying fire blanket” system, using a team of relatively simple and cheap unmanned aerial vehicles to collaboratively tether a fire blanket to be dropped at a designated place. This system puts a high demand on the safe, precise, and resilient operations of the aerial vehicle team. The intellectual merits of the project include new adaptive and learning cooperative formation architecture designs that will significantly advance the current state-of-the-art in cooperative control algorithms. The broader impacts of the project include collaboration with the industry, our Lexington Fire Department, and Kentucky Division of Forestry, on the designs and verifications of the system, and on promoting the system for real-world fire fighting and prevention. This project will advance research and educational experiences for K-12 and undergraduate students including underrepresented students. A graduate course on intelligent control methods will be developed based on the research findings of this project. The goal of this project is to develop new adaptive and learning architectures for physically interconnected unmanned aerial vehicles to ensure safe, precise, and resilient operations. Major technical challenges to be addressed include: (1) satisfying multiple formation constraint requirements that are time and path dependent; (2) adaptation and learning under varying operation conditions over both the time and iteration domain; and (3) resilient designs in the face of potential malfunctioning at the fire scene. Current cooperative control algorithms mostly focus on constant or time-varying constraint requirements, which often require more aggressive kinematic behavior that can potentially cause actuation saturation. The geometric and spatial nature of the constraint requirements is often overlooked, largely due to difficulties in integrating time-domain system kinematics with path-domain constraints. Moreover, existing works can at best deal with adaption or learning over either the time or iteration axis. Unified structures to learn and adapt over both the time and iteration domain have not been addressed in the literature. Furthermore, existing cooperative formation algorithms often ignore any malfunctioning of the physically interconnected agents during constrained operations. To address these challenges, we will use barrier Lyapunov analysis and a new framework of composite energy function discussion, to develop new adaptive and learning cooperative formation architectures based on universal barrier functions, adaptive learning structures, and resilient control framework designs, while taking unknown external payloads, system nonlinearities, and external disturbances into considerations.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.