CPS: Synergy: Collaborative Research: Computationally Aware Cyber-Physical Systems

CPS：协同：协作研究：计算感知网络物理系统

• 批准号: 1544396
• 负责人: Ricardo Sanfelice
• 金额: $ 43.2万
• 依托单位: University of California-Santa Cruz
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2015
• 资助国家: 美国
• 起止时间: 2015-09-15 至 2020-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1544396&HistoricalAwards=false
• 关键词: CPS; Synergy; Collaborative; Research; Computationally

The objective of this work is to generate new fundamental science that enables the operation of cyber-physical systems through complex environments. Predicting how a system will behave in the future requires more computing power if that system is complex. Navigating through environments with many obstacles could require significant computing time, which may delay the issue of decisions that have to be made by the on-board algorithms. Fortunately, systems do not always need the most accurate model to predict their behavior. This project develops new theory for deciding between the best model to use when making a decision in real time. The approach involves switching between different predictive models of the system, depending on the computational burden of the associated controller, and the accuracy that the predictive model provides. These tools will pave the way for more kinds of aircraft to navigate closely and safely with one another through the National Air Space (NAS), including Unmanned Air Systems (UAS). 

The results from this project will enable more accurate and faster trajectory synthesis for controllers with nonlinear plants, or nonlinear constraints that encode obstacles. The approach utilizes hybrid control to switch between models whose accuracy is normalized by their computational burden of predictive control methods. This synergistic approach enables computationally-aware cyber-physical systems (CPSs), in which model accuracy can be jointly considered with computational requirements. The project advances the knowledge on modeling, analysis, and design of CPSs that utilize predictive methods for trajectory synthesis under constraints in real-time cyber-physical systems. 
The results will include methods for the design of algorithms that adapt to the computational limitations of autonomous and semi-autonomous systems while satisfying stringent timing and safety requirements. With these methods come new tools to account for computational capabilities in real-time, and new hybrid feedback algorithms and prediction schemes that exploit computational capabilities to arrive at more accurate predictions within the time constraints. The algorithms will be modeled in terms of hybrid dynamical systems, to guarantee dynamical properties of interest. The problem space will draw from models of UAS in the NAS.

这项工作的目标是产生新的基础科学，使网络物理系统能够在复杂的环境中运行。如果一个系统很复杂，预测该系统未来的行为需要更多的计算能力。在有许多障碍物的环境中导航可能需要大量的计算时间，这可能会延迟必须由机载算法做出的决策。幸运的是，系统并不总是需要最准确的模型来预测其行为。该项目开发了新的理论，用于在真实的时间内做出决策时使用最佳模型。该方法涉及在系统的不同预测模型之间切换，这取决于相关联的控制器的计算负担以及预测模型提供的精度。这些工具将为更多种类的飞机通过国家空域（NAS），包括无人机系统（UAS），相互密切安全地导航铺平道路。

该项目的结果将使具有非线性对象或编码障碍物的非线性约束的控制器能够更准确和更快地进行轨迹合成。该方法利用混合控制模型之间的切换，其准确性是归一化的预测控制方法的计算负担。这种协同方法使计算感知的网络物理系统（CPS），其中模型精度可以与计算要求共同考虑。该项目推进了CPS建模、分析和设计方面的知识，这些CPS利用预测方法在实时网络物理系统的约束下进行轨迹合成。
结果将包括算法的设计方法，适应自主和半自主系统的计算限制，同时满足严格的时序和安全要求。随着这些方法的出现，新的工具可以实时计算能力，新的混合反馈算法和预测方案可以利用计算能力在时间限制内获得更准确的预测。该算法将建模的混合动力系统，以保证感兴趣的动态特性。问题空间将从NAS中的UAS模型中提取。