CPS: Breakthrough: Solar-powered, Long-endurance UAV for Real-time Onboard Data Processing

近年来，无人驾驶飞行器(UAV)的普及有了实质性的上升趋势。这些飞机在多个领域得到应用，如精准农业、基础设施和环境监测、监视、测绘、搜索和救援任务、紧急情况和自然灾害的快速评估、下一代互联网连接、天气判断等。考虑到各种可能性，无人机代表着CPS的一个不断增长的市场，它们被视为一种“使能技术”，可以在全球范围内重新考虑人类参与许多军事和民用应用的问题。实现这一增长的主要挑战之一是无人机的耐力。这与无人机执行任务所需的能量直接相关。这项提议旨在通过在无人机性能和节能计算之间进行权衡来增加无人机的耐力。这需要将任务和目标映射到能源需求和计算需求。该项目的目标是表明，这种贸易可以实现长时间飞行，特别是在太阳能被用作一次能源的情况下。这一雄心勃勃的计划是开发一种重量轻、效率高的飞机，能够进行机动感知功率适应和实时视频/传感器采集和处理，最长可持续飞行12小时(这一限制由白天设定)。该项目旨在扩大无人机设计和集成的理论和实践基础，该无人机能够从视觉、声音和其他传感器阵列进行实时感知和处理。小型无人机的传统方法是捕获飞机上的数据，通过高功率数据链路将其传输到地面，远程处理，执行分析，然后根据需要将命令传递回飞机。相反，这项研究的目标是具有零碳足迹的太阳能无人机，该无人机携带能够在运行时预算推进/驱动子系统与计算和通信子系统之间的可用功率的高性能嵌入式计算机系统有效负载。首先，将为所考虑的无人机构建一组准确的功率模型，以建立不同飞行模式(飞机机动)和推进/驱动子系统相应的功率需求之间的映射。其次，将开发软件和硬件级的功率自适应机制，以设计一种适用于无人机的新型功率自适应集成模块化航空电子(PA-IMA)体系结构。应用程序之间安全的时间/空间分区和灵活的调度以处理飞行中不可预测的功率/负载变化是关键需求。一旦获得飞行和计算模式的准确表征，将开发更高级别的监督逻辑，以在推进/驱动子系统和计算/通信子系统之间分配可用功率预算。虽然精准农业和土地/基础设施监测将立即受益于这种技术，但这项研究的长期影响要广泛得多，因为它探索了环境感知能力和计算管理的基础。总体而言，开发的理论将适用于功率预算有限且可变的自动驾驶车辆和机器人：这些都是收集太阳能和风能时面临的常见挑战。

In recent years, there has been a substantial uptrend in the popularity of unmanned aerial vehicles (UAVs). These aircraft find application in several areas such as precision farming, infrastructure and environment monitoring, surveillance, surveying and mapping, search and rescue missions, rapid assessment of emergency situations and natural disasters, next generation Internet connectivity, weather determination and more. Given the wide range of possibilities, UAVs represent a growing market in CPS and they are perceived as an "enabling technology" to re-consider the human involvement in many military and civil applications on a global scale. One of the major challenges in enabling this growth is UAV endurance. This is directly related to the amount of energy available to the UAV to perform its mission. This proposal looks to increase UAV endurance by trading off UAV performance with energy efficient computing. This requires mapping of mission and goals into energy needs and computational requirements. The goal of the project is to show that this trade can enable long-duration flight especially when solar energy is utilized as a primary energy source. The ambitious plan is to develop a light weight and efficient aircraft capable of maneuver-aware power adaptation and real-time video/sensor acquisition and processing for up to 12 hours of continuous flight (this limit being set by daylight hours).This project aims to expanding the theoretical and practical foundations for the design and integration of UAVs capable of real-time sensing and processing from an array of visual, acoustic and other sensors. The traditional approach for small size UAVs is to capture data on the aircraft, stream it to the ground through a high power data-link, process it remotely, perform analysis, and then relay commands back to the aircraft as needed. Conversely, this research targets a solar-powered UAV with a zero-carbon footprint that carries a high performance embedded computer system payload capable of budgeting at run-time the available power between the propulsion/actuation subsystems and the computing and communication subsystems. First, a set of accurate power models for the considered UAV will be constructed to establish a mapping between different flight modes (aircraft maneuvers) and the corresponding power requirements at the propulsion/actuation subsystem. Second, software and hardware-level power adaptation mechanisms will be developed to devise a novel Power Adaptive Integrated Modular Avionic (PA-IMA) architecture suitable for UAVs. Safe temporal/spatial partitioning among applications and flexible scheduling to handle unpredictable power/load variations in flight represent key requirements. Once an accurate characterization is available for flight and computation modes, a higher-level supervisory logic will be developed to distribute the available power budget between the propulsion/actuation subsystem and the computation/communication subsystem. While precision farming and land/infrastructure monitoring will immediately benefit from such a technology, the long-term impact of this research is much broader since it explores the very foundations of environment-aware power and computation management. In general, the developed theory will be applicable to autonomous vehicles and robots whose power budget is limited and variable: these are common challenges faced when harvesting solar and wind energy.