CRII: CPS: Towards Optimal Information Gathering in Unknown Stochastic Environments

CRII：CPS：在未知随机环境中实现最佳信息收集

• 批准号: 1566240
• 负责人: Zak Kassas
• 金额: $ 17.5万
• 依托单位: University of California-Riverside
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2016
• 资助国家: 美国
• 起止时间: 2016-05-01 至 2019-05-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1566240&HistoricalAwards=false
• 关键词: CRII; CPS; Towards; Optimal; Information

Society is steadily moving toward a world in which autonomous vehicles roam the streets side-by-side with human-driven vehicles and unmanned aerial vehicles are integrated into the national airspace. These autonomous systems will be tasked with missions such as search, rescue, surveillance, reconnaissance, mapping, farming, fire fighting, and transportation. Future autonomous systems will operate in unfamiliar areas with minimal or no human interaction for prolonged periods of time. The luxury of building prior detailed maps of these environments could be (1) prohibitive (e.g., disaster areas), (2) impractical (e.g., signal landscapes and congested downtowns), or (3) economically not viable (e.g., hospital buildings and national forests). With no human-in-the-loop before or during operation, one expects future autonomous systems to (1) possess full situational awareness and (2) gather sufficient information about their environment. These two tasks need to seamlessly integrate into the overall mission of the autonomous system. Current autonomous systems are far from possessing these capabilities, and the current analytical tools are insufficient to deal with this emerging class of problems.This project will develop a coherent analytical foundation and a suite of algorithms and tools for autonomous systems deployed in unknown, dynamic stochastic environments to optimally gather sufficient information to successfully accomplish their mission. The research specifically considers autonomous systems with limited sensing, computation, actuation, and communication capabilities. This research will study a new class of information optimization measures, which possess desirable convexity properties (enabling real-time execution) and separability properties (enabling near-lossless distributed implementation among agents). This research aims to establish fundamental relationships between performance and computational complexity in the presence of varying degrees of environmental uncertainty. These relationships will enable principled navigation of these complex trade-offs, leading to autonomous identification and adoption of the optimal information gathering strategy. This project has a vertically-integrated education plan spanning K-12, undergraduate, and graduate students. The project will also train in-service and pre-service K-12 teachers to apply Next Generation Science Standards (NGSS) - a set of science standards that integrate rigorous content and application, reflecting how STEM is practiced in the real world. This research has far-reaching impact - it will evolve autonomous systems from sensing the environment to making sense of the environment, bringing new capabilities in environments where direct human control is physically or economically not possible.

社会正在稳步迈向这样一个世界：自动驾驶汽车与人类驾驶的汽车并肩行驶在街道上，无人机融入国家领空。这些自主系统将承担搜索、救援、监视、侦察、测绘、农业、消防和运输等任务。未来的自主系统将在不熟悉的区域运行，在很长一段时间内很少或没有人类互动。预先构建这些环境的详细地图可能会 (1) 令人望而却步（例如，灾区），(2) 不切实际（例如，信号景观和拥挤的市中心），或 (3) 经济上不可行（例如，医院建筑和国家森林）。由于在操作之前或操作过程中没有人参与，人们期望未来的自主系统能够（1）拥有完整的态势感知能力，（2）收集有关其环境的足够信息。这两项任务需要无缝集成到自治系统的整体任务中。当前的自主系统远未具备这些能力，并且当前的分析工具不足以处理这类新兴问题。该项目将为部署在未知的动态随机环境中的自主系统开发一个连贯的分析基础和一套算法和工具。环境以最佳方式收集足够的信息以成功完成其任务。该研究特别考虑了传感、计算、驱动和通信能力有限的自主系统。本研究将研究一类新型信息优化措施，其具有理想的凸性特性（实现实时执行）和可分离性特性（实现代理之间近乎无损的分布式实施）。本研究旨在在存在不同程度的环境不确定性的情况下建立性能和计算复杂性之间的基本关系。这些关系将使这些复杂的权衡得以原则性导航，从而实现自主识别和采用最佳信息收集策略。该项目有一个涵盖 K-12、本科生和研究生的垂直整合教育计划。该项目还将培训在职和岗前 K-12 教师应用下一代科学标准 (NGSS)，这是一套将严格的内容和应用相结合的科学标准，反映了 STEM 在现实世界中的实践方式。这项研究具有深远的影响——它将把自主系统从感知环境发展到理解环境，为人类在物理上或经济上不可能直接控制的环境带来新的能力。