Robust and Optimal Control of Physical Systems with Networked Communication Structures

毫无疑问，我们现在生活在一个网络化的世界里。许多物理、虚拟或混合（物理和虚拟的混合）系统通过不同类型的网络结构（无线或有线）进行操作、通信或控制。例如：互联网系统、电网、汽车系统、各种传感器网络等。这些网络化系统的共同性质是，它们既表现为动态系统又表现为网络，即同时具有动态系统和网络的特征，这既带来了优势（如降低布线和维护成本，增强系统和网络层面的灵活功能等），也带来了挑战（如增加系统或系统组件之间交互的复杂性）。由于网络通信等原因，对时间敏感的性能折衷)。例如，在优点方面，通过采用控制器局域网（CAN）总线用于发动机控制等关键应用，以及采用本地互联网络（LIN）总线用于动力窗/座椅控制等非关键应用，人们看到现在每辆汽车的布线成本大大降低；另一方面，在挑战方面，人们已经意识到，分布式传感器、控制器和底层植物系统之间的数据网络通信往往会降低被控系统的瞬态动态性能，如果不仔细处理，可能会导致整个系统的不稳定。已经确定，通信网络带来的挑战是许多实际时间敏感的工业系统，如电网，汽车发动机控制系统和传感器网络等的极大兴趣。尽管这一挑战在最近发表的大量文献中受到了攻击，但许多重要的问题仍然没有解决，例如，到目前为止，人们仍然不知道如何设计一个控制器，它可以同时处理网络通信通道中的量化和噪声效应，并调节被控目标系统的各种期望的动态行为。此外，一些重要的问题，如网络通信对多智能体系统的性能意味着什么，网络通信如何影响分布式控制系统的估计性能，以及网络通信如何影响具有时间和事件触发机制的控制系统的性能等，仍然没有触及，但这些问题在各个行业中都是“最想要”的问题。在这项发现拨款提案中，重点将放在开发系统设计工具上，以解决具有网络通信结构的物理控制系统的时间敏感性能。特别是，本文认为信息论和控制论应该结合起来，产生新的概念方法来解决上述问题，并为网络动力系统设计鲁棒和最优的控制和管理机制。还有人认为，所开发的方法将以工业需要为导向，因此将在实践中找到友好的应用方式。

There is no doubt that we are in a networked world right now. Many physical, virtual, or hybrid (mixed physical and virtual) systems operate, communicate, or are controlled through different kinds of network structures, wireless or wired. Examples can be given as: internet systems, electric power grids, automotive systems, various sensor networks, etc. The common nature of these networked systems is that they are exhibited as both dynamical systems and networks, i.e., carrying the features of both dynamical systems and networks which brings in both advantages (such as cost reduction for wiring and maintenance, enhanced flexible functions at both system and network levels, etc.) and challenges (such as increased complexities in the interaction among systems or system components, time-sensitive performance compromise due to networked communication, etc. ). For example, on the advantage side, people have seen tremendous reduction of the wiring costs in every automobile nowadays by adopting the Controller Area Network (CAN) bus for critical applications such as engine control and the Local Interconnected Network (LIN) bus for noncritical applications such as power window/seat control; on the challenge aspect, on the other hand, it has already been realized that the networked communication of data among distributed sensors, controllers, and underlying plant systems tends to downgrade the transient dynamical performance of the controlled systems which, if not carefully addressed, could result in the instability of the whole systems. It has been identified that the challenge posed by the communication network is of great interest to many practical time-sensitive industrial systems such as electric power grids, automotive engine control systems, and sensor network, etc. Although this challenge has been attacked in great number of published literature most recently, many important questions remain open, for example, so far one still does not know how to design a controller that could simultaneously address quantizing and noise effects in the network communication channels and regulate the various desired dynamical behaviors of the controlled target systems. Also, some important questions, such as what the networked communication means to the performance of multi-agent systems, how the networked communication would affect the performance of estimation in distributed control systems, and how the networked communication affect the performance of control systems with both time and event-triggered mechanism, etc., still remain untouched, yet these questions are on top of the ‘most-wanted’ list in various industries. In this Discovery Grant proposal, the focus will be on developing systematic design tools to address the time-sensitive performance in physical control systems with networked communication structure. In particular, it is argued that the information theory and control theory should be integrated to generate conceptually new methods to address the said problems and to design robust and optimal control and management mechanisms for the networked dynamical systems. It is also argued that the methods developed would be industrial needs oriented and hence would find their way to be friendly applied in practice.