CSR-EHCS(EHS), SM: Collaborative Research: An Anytime Approach to Real-Time Embedded Control

网络物理系统将很快变得无处不在。这种系统带来的主要挑战之一是，传统上，调节与物理世界交互的控制算法假设无限计算资源的可用性。然而，在网络物理系统中，控制任务是在共享处理器上执行的，这些处理器只能提供时变和不确定的计算资源。此外，考虑到控制性能对计算资源时变可用性引起的抖动和延迟的敏感性，控制任务的实时调度给这种情况带来了新的困难。这些都是存在于控制算法和实时调度策略之间的新问题。该研究提供了一种新的联合控制和任务调度方法，在时变处理器可用性下保持保证的控制性能，同时最优地利用可用的计算资源。更具体地说，本项目旨在设计新颖的随时后退地平线控制算法，随着提供的计算时间的增加，算法的性能逐渐提高，并根据控制性能和其他实时任务的服务质量在线重新分配可用的计算资源。通过解决网络物理系统设计中的一个基本瓶颈，该项目将以深刻的方式直接影响社会。视距后退控制在实时环境中的新型应用，如基于网络视觉的控制和复杂的SCADA系统，将具有经济和社会效益。该项目还将注意力集中在控制和实时计算交叉的挑战上，从而促进了这两个社区研究人员之间的合作，也带来了新的教育材料。

Cyber-physical systems will soon become ubiquitous. One of the major challenges that such systems pose is that the control algorithms that modulate the interaction with the physical world have traditionally assumed availability of unlimited computational resources. However, in cyber-physical systems, control tasks are executed on shared processors that can only provide time-varying and uncertain computational resources. Moreover, the real-time scheduling of control tasks provides new difficulties in this context given the sensitivity of control performance to jitter and latency caused by the time varying availability of computational resources. These are new problems that reside at the interface between control algorithms and real-time scheduling policies. This research provides a new joint control and task scheduling approach that maintaining a guaranteed control performance under time-varying processor availability while optimally utilizing the available computational resources. More specifically, this project aims at the design of novel anytime receding horizon control algorithms that provide progressively better performance as more computational time is provided and redistribute the available computational resources online based on the control performance and on the quality of service of other real-time tasks.By addressing a fundamental bottleneck in the design of cyber-physical systems, this project directly impacts the society in profound ways. New classes of applications of receding horizon control in real-time environments, such as networked vision-based control and complex SCADA systems will become possible with economic and societal benefits. This project also focuses attention at the challenges at the intersection of control and real-time computation, and thus fosters collaboration between researchers in these two communities, also leading to new educational material.