CPS: Breakthrough: Programming and Execution Environment for Geo-Distributed Latency-Sensitive Applications

新的网络传感技术（如摄像头）、分布式计算资源（如云计算）和算法进步（如计算机视觉）的融合为解决各种具有社会重要性的新问题提供了令人兴奋的新机会，这些新问题包括应急响应、灾难恢复、监视和运输。这类新问题的解决方案被称为“态势感知”应用，包括通过大规模分布式摄像头网络进行监控，以及使用道路和交通传感的车辆网络中的个性化交通警报。需要在系统软件技术方面取得突破，以应对这些应用程序带来的挑战，因为它们对延迟敏感、数据密集、涉及重型处理，并且必须在处理物理世界的变幻莫测时全天候运行。本项目旨在通过新的分布式编程习惯用法和资源分配策略实现这一突破。为了更好地识别态势感知应用带来的挑战，该项目包括与路易斯安那州巴吞鲁日市合作进行新技术的实验部署。核心活动是为态势感知应用程序的设计开发适当的系统抽象，并将它们封装在领域专家（例如视觉研究人员）的分布式编程习惯中。生成的编程框架允许将关键属性（如位置、时间和移动性）与感知数据关联起来，从而沿着这些轴推断因果事件。为了满足这些应用程序的延迟限制，该项目开发了地理空间资源分配机制，以补充和支持分布式编程习惯，将云计算的效用计算模型扩展到网络边缘。由于分布式传感源的限制，应用程序通常不得不与物理环境中正在发生的事情的不精确知识一起工作，因此该项目还研究了特定应用程序信息融合和时空分析的系统支持，以提高结果的质量。发展未来网络物理系统劳动力的努力包括围绕态势感知创建新的多学科课程，探索新的沉浸式学习教学风格，以及通过研究经验和实习为本科生提供指导和研究经验，旨在增加女性和少数民族的参与。

The confluence of new networked sensing technologies (e.g., cameras), distributed computational resources (e.g., cloud computing), and algorithmic advances (e.g., computer vision) are offering new and exciting opportunities for solving a variety of new problems that are of societal importance including emergency response, disaster recovery, surveillance, and transportation. Solutions to this new class of problems, referred to as "situation awareness" applications, include surveillance via large-scale distributed camera networks and personalized traffic alerts in vehicular networks using road and traffic sensing. A breakthrough in system software technology is needed to meet the challenges posed by these applications since they are latency-sensitive, data intensive, involve heavy-duty processing, and must run 24x7 while dealing with the vagaries of the physical world. This project aims to make such a breakthrough, through new distributed programming idioms and resource allocation strategies. To better identify the challenges posed by situation awareness applications, the project includes experimental deployment of the new technologies in partnership with the City of Baton Rouge, Louisiana. The central activity is to develop appropriate system abstractions for design of situation awareness applications and encapsulate them in distributed programming idioms for domain experts (e.g., vision researchers). The resulting programming framework allows association of critical attributes such as location, time, and mobility with sensed data to reason about causal events along these axes. To meet the latency constraints of these applications, the project develops geospatial resource allocation mechanisms that complement and support the distributed programming idioms, extending the utility-computing model of cloud computing to the edge of the network. Since the applications often have to work with inexact knowledge of what is happening in the physical environment, owing to limitations of the distributed sensing sources, the project also investigates system support for application-specific information fusion and spatio-temporal analyses to increase the quality of results. Efforts toward development of a future cyber-physical systems workforce include creation of a new multidisciplinary curriculum around situation awareness, exploration of new immersive learning pedagogical styles, and mentoring and providing research experience to undergraduate students through research experiences and internships aimed at increasing participation of women and minorities.