ERI: Fault-Tolerant Monitoring of Moving Clusters of Targets using Collaborative Unmanned Aerial Vehicles

无人驾驶飞行器或无人机已成功用于监测地面活动。然而，目前还不可能长时间使用小型无人机，因此限制了它们的实施。例如，易于运输和部署的小型四轴飞行器在大多数情况下飞行时间不超过40分钟，并且容易受到意外故障的影响，如自然灾害造成的损坏。另一方面，能够延长飞行时间的坚固的四轴飞行器的尺寸和重量都较大，因此无法轻松部署。作为单一坚固无人机的替代方案，这一工程研究启动(ERI)奖将支持基础研究，使小型无人机网络能够监测地面活动，目标是通过相互共享信息，包括检测到的目标的知识，不间断地运行和容错。这一概念将通过与美国林业局合作的野火监测进行演示。这一奖项将支持一个以本科生为主的机构的研究。研究中考虑的监控问题与著名的多旅行商问题及其变种有关，即带时间窗的车辆路径问题和多仓库无人机路径问题。然而，这些路线问题的解决方案不能按原样使用，因为他们会考虑只访问一个集群(地区)一次的无人机旅行，而不是定期访问。此外，它们没有考虑容错。该研究旨在提出一种容错解决方案：(1)通过高斯混合模型的分布式估计来表征目标簇；(2)利用博弈论协调飞行编队和搜索路径以避免中央控制；(3)对来自无人机的相邻图像进行点集配准，以提高目标定位的精度。这项研究不仅将促进野火监测方面的进展，而且还将促进任何其他可以以高斯混合模型为特征的自然或人类活动。该项目由跨部门机器人基础研究计划支持，该计划由工程指导委员会(ENG)和计算机和信息科学与工程指导委员会(CEISE)联合管理和资助。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Unmanned aerial vehicles, or drones, have successfully been used to monitor ground activity. However, using small drones for extended periods of time is not yet possible, thus limiting their implementation. For instance, small quadcopters that can be easily transported and deployed do not exceed forty minutes of flying time in most cases and are susceptible to unexpected failure such as damage from natural hazards. On the other hand, robust quadcopters capable of longer flying times have larger dimensions and weight that prohibit ease of deployment. As an alternative to a single robust drone, this Engineering Research Initiation (ERI) award will support fundamental research to enable a network of small drones to monitor ground activity with the goal of uninterrupted operation and fault tolerance by sharing information with one another including knowledge of targets detected. A demonstration of this concept will be made through wildfire monitoring in collaboration with the US Forest Service. This award will sustain research at a predominantly undergraduate institution. Both undergraduate and graduate students will participate in the research effort.The monitoring problem under consideration in this research is related to the well-known multiple traveling salesman problem and its variants, namely Vehicle Routing Problem with Time Window and Multiple Depot Drone Routing Problem. However, the solution to these routing problems cannot be used as-is because they would consider drone tours that visit each cluster (region) only once, not periodically. Moreover, they do not consider fault tolerance. This research aims at a fault-tolerant solution that (1) characterizes target clusters via distributed estimation of Gaussian Mixture Models, (2) coordinates flight formations and search paths using game theory to avoid central control, and (3) performs point set registration of adjacent images from drones to increase accuracy of target locations. The research will not only promote progress in wildfire monitoring but also for any other natural or human activity that can be characterized with Gaussian Mixture Models.This project is supported by the cross-directorate Foundational Research in Robotics program, jointly managed and funded by the Directorates for Engineering (ENG) and Computer and Information Science and Engineering (CISE).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.