S&AS: FND: Safe Task-Aware Autonomous Resilient Systems (STAARS)

实现无人驾驶航空系统(UAS)的商业和社会效益的全部潜力，将要求自主UAS必须围绕人，特别是城市/郊区，并尊重安全、隐私和监管方面的考虑。该项目将把无人机在城市/郊区环境中执行任务的操作量化为这些考虑因素的“风险”，然后开发动态风险评估和指导算法，以计算出无人机在执行任务时遵循的风险最低的轨迹。该项目将产生关于UAS在城市运营中的适当自治水平的知识，并将在UAS监管问题上有利于联邦航空局(FAA)。该项目的技术进步还将为包括自主、移动网络和智能控制在内的多个领域做出贡献。该项目开发的任务和风险感知决策框架不仅可以应用于无人机，还可以应用于地面和水下的其他无人系统，在智能医疗、交通运输和制造领域具有广泛的应用。投资推广署亦预期，成功示范在不同风险情况下“安全”操作UAS，将会增加公众对UAS技术的接受程度，并惠及广泛的商用UAS用途和就业市场。该项目还将为学生和整个社区提供令人兴奋的学习和培训机会，以学习无人机技术。该项目将使用两个栅格地图来量化风险：(1)PREM(概率风险暴露地图)定义了地面人员和财产暴露于空中无人机的风险，作为位置和时间的函数。(2)PURM(概率UAS可达性图)是UAS从其当前位置到达地面位置的概率，基于车辆的能力(包括标称能力和可能发生故障而减弱的能力)和环境条件(如风)计算。通过名义和所有故障情况的可达性域的联合概率来定义PURM会导致弹性系统，因为决策是考虑所有可能的操作模式做出的。PURM中的轨迹规划将使用改进的双向概率RRT(快速扩展随机树)来高效、增量地规划一组轨迹，从而将总体风险降至最低。然后，自主决策算法可以将风险保持在可接受的水平以下，因为它指导UAS成功完成给定任务。由于无人机的安全运行还高度依赖于无人机之间以及无人机与指挥控制中心之间的有效通信，因此该项目还将制定不同风险水平和机动性约束下的分散式动态通信方案。

Realizing the full potential of unmanned aerial systems (UAS) for commercial and societal benefits will call for autonomous UAS that must operate around people, especially urban/suburban areas, and respect safety, privacy, and regulatory concerns. This project will quantify the operations of UAS executing a task in urban/suburban environment as "risk" to these considerations, then develop dynamic risk assessment and guidance algorithms to compute least risky trajectories for the UAS to follow while executing a task. The project will produce knowledge on the proper autonomy level of UAS in urban operations, and will benefit FAA (Federal Aviation Administration) in UAS regulatory issues. The technological advances in this project will also contribute to multiple fields including autonomy, mobile networking, and intelligent control. The task & risk-aware decision-making framework developed in this project can be applied not only to UAS, but also to other unmanned systems on the ground and in/under water, with broad applications in smart health, transportation, and manufacturing domains. The PIs also expect that successful demonstrations of "safe" UAS operation in various risk conditions will increase the public's acceptance of UAS technology, and benefit broad commercial UAS use and job market. The project will also produce exciting learning and training opportunities for students and the community at large to learn UAS technologies.The project will use two raster maps to quantify risk: (1) PREM (Probability Risk Exposure Map) defines the risk of exposure of people and property on the ground to the presence of a UAS in the air as a function of position and time. (2) PURM (Probabilistic UAS Reachability Map) is the probability that the UAS can reach a position on the ground from its current position, computed based on the vehicle's capability (both nominal and diminished by possible failures) and environmental conditions such as wind. Defining the PURM by joint probability of reachability domains for nominal and all failure cases results in a resilient system, since decisions are made considering all possible operational modes. Trajectory planning in the PURM will use a modified, bidirectional, probabilistic RRT (Rapidly Expanding Random Tree) to efficiently, incrementally plan a set of trajectories that minimizes the overall risk. An autonomous decision algorithm then can keep the risk below an acceptable level as it guides the UAS in the successful completion of a given task. Because the safe operation of UAS also highly relies on effective communication among UAS and between UAS and a Command and Control Center, the project will also develop a decentralized dynamic communication schemes under different risk levels and mobility constraints.