Scalable system level synthesis for virtually localizable systems
Scalable system level synthesis for virtually localizable systems
复制标题
用于几乎可本地化系统的可扩展系统级综合
DOI：
10.1109/cdc.2017.8264168
复制
发表时间：
2017
期刊：
影响因子：
--
通讯作者：
James Anderson
中科院分区：
文献类型：
--
作者：
N. Matni;Yuh;James Anderson
在以前的工作中，我们开发了系统级的控制器综合方法，并表明，在适当的假设下，这个框架允许本地化控制器的合成。我们进一步表明，这种本地化的控制器享有O（1）的合成和实现复杂度相对于全球系统的尺寸，使他们特别适合于控制大规模的网络物理系统。然而，一个系统是本地化的假设是严格的：粗略地说，一个系统是本地化的，如果控制器有必要的驱动，传感和通信资源，以“提前”的干扰传播和中和它，从而包含其影响到本地化的时空区域。在本文中，我们放松了假设的精确定位，并开发了一种控制器的综合方法，适用于任意系统，在适当的意义上“容易控制”。我们专注于状态反馈设置和发展一个简单的必要条件和充分条件的鲁棒稳定性使用系统级的方法。然后，我们利用这个条件，沿着引入虚拟驱动，通信和系统响应到合成过程中，设计稳定控制器，具有（i）O（1）合成和实现的复杂性和（ii）保证性能界限。最后，我们的电源启发的例子证明了这些技术的有用性，其中我们合成一个接近全局最优控制器的系统，既不是本地化的，也不是二次不变的。
In previous work, we developed the system level approach to controller synthesis, and showed that under suitable assumptions, this framework allowed for the synthesis of localized controllers. We further showed that such localized controllers enjoy O(1) synthesis and implementation complexity relative to the dimension of the global system, making them particularly well suited for the control of large-scale cyber-physical systems. However, the assumptions under which a system is localizable are stringent: roughly, a system is localizable if the controller has the necessary actuation, sensing and communication resources to “get out ahead” of the propagation of a disturbance and neutralize it, thus containing its effect to a localized spatiotemporal region. In this paper, we relax the assumption of exact localizability, and develop a controller synthesis methodology that is applicable to arbitrary systems that are in an appropriate sense “easy to control.” We focus on the state-feedback setting and develop a simple necessary and sufficient condition for robust stability using the system level approach. We then leverage this condition, along with the introduction of virtual actuation, communication and system responses into the synthesis process, to design stabilizing controllers that have (i) O(1) synthesis and implementation complexity and (ii) guaranteed performance bounds. We end with a power-inspired example demonstrating the usefulness of these techniques, wherein we synthesize a near globally optimal controller for a system that is neither localizable nor quadratically invariant.