CSR---CPS: Design of Robust and Energy Efficient Cyber-Physical Systems Using Dynamical Systems and Control Theory

这项研究致力于利用动力系统理论、控制理论和计算机科学的工具，开发用于建立网络物理系统的动力学模型和支持健壮性分析的形式化方法。具体地说，该项目解决了开发有效的不确定性传播技术的需要，这种不确定性传播是由于时变任务集引起的操作不确定性和网络物理系统中存在的控制障碍。其方法是将网络物理系统抽象为具有异质体系结构平台和多功能范型的分布式嵌入式系统。这种抽象是在运行时和存在不确定性的情况下利用分配的任务来完成实时、多模式操作所必需的。考虑的操作不确定性包括由操作模式改变引起的不同任务负载、部署在恶劣环境中的一些元件的故障、瞬时计算过载和通信延迟。研究的关键技术包括基于遍历理论、面向集合的数值技术、动力学的多尺度马尔可夫链近似和谱图划分与聚集概念的嵌入式系统不确定性管理方法；使用抽样方法和有限时间估计技术对网络物理任务进行建模；以及为网络物理系统的稳健性能提供李雅普诺夫证书。在系统开发中使用这样的框架可望显著提高网络物理系统的可靠性。

This research effort is focused on developing formal methods for modeling dynamics and supporting robustness analysis of cyber-physical systems, using tools from dynamical systems theory, control theory and computer science. Specifically, the project addresses the need to develop efficient techniques for uncertainty propagation that exists due to operational uncertainties arising from time-varying task sets and control impediments present in cyber-physical systems. The approach is to abstract cyber-physical systems as distributed embedded systems with heterogeneous architecture platforms and multi-functional paradigms. Such abstractions are required to accomplish real-time, multi-mode operations with tasks allocated at run-time, and in the presence of uncertainty. The operational uncertainties considered include those due to varying task loads caused by operational mode changes, the failure of a few elements when deployed in a hostile environment, transient computational overloads and communication delays. Key technical thrusts in this research include developing methods for uncertainty management in embedded systems based on methods of ergodic theory, set-oriented numerical techniques, multi-scale Markov chain approximations of dynamics and spectral graph partitioning and aggregation concepts; modeling of cyber-physical tasks using sampling methods and finite horizon estimation techniques; and providing Lyapunov certificates for robust performance of cyber-physical systems. Using such a framework in system development is expected to significantly improve the reliability of cyber-physical systems.