CAREER: Hybrid Control of Complex Networked Systems

本项目致力于混合系统理论领域的研究和教学。目标是为复杂的系统网络开发一种控制理论，每个系统都有自己的目标、决策和信息处理能力。系统的动力学主要是连续状态和连续时间的，但是聚合网络通常过于复杂，无法仅用连续工具进行建模。迄今为止，控制理论在单个控制系统的分析和综合以及通过点对点链路连接在一起以使每个子系统同步接收和处理信息的系统群的控制规律的发展方面取得了巨大的成功。使用连续状态工具，也设计了简单的分层控制方案，并取得了巨大的成功。飞机的控制就是一个很好的例子：控制当今商用喷气式飞机表面的复杂液压阀系统对飞行员来说是透明的，因为在两者之间有一个控制层，它限制了液压系统的操作，并向飞行员显示了复杂系统的简化抽象。我们现在感兴趣的是为比连续、同步控制理论所允许的复杂得多的系统设计控制律。例子包括空中交通管制系统、高速公路系统、通信系统、互联电网、金融网络（如股票市场）和分布式MEMS传感器网络。在这些示例中，传感器和资源分布在各个系统中，但是系统非常需要协调以执行某些任务或实现某些目标。系统也是异步的，因为控制器在不同的时间可以获得相关但不同的信息。我们建议设计分层混合控制系统：我们的控制系统将是一个分层结构，其中最高层使用离散状态模型来模拟系统中的大模式空间，最低层包含分布式传感和控制，信息的异步接收和处理以及多目标函数。我们的研究将集中在：1。实时混合接口的设计：本研究的重点是在离散状态系统抽象和连续状态子系统之间开发一个鲁棒且计算效率高的混合系统接口。该接口将提供先验证明，证明系统将按预期运行，并自动综合离散和连续控制律。异步系统的控制：与自主交互代理一起出现的是异步问题。在经典控制理论中，对传感器数据进行采样，并以固定速率发出执行器命令。我们建议设计一种异步控制理论，其中传感器在有新信息时传输带有时间戳的数据包，控制器接收与其当前功能相关的数据包。自动化空中交通系统的算法：在未来的空中运输系统中，建议将当前的空中交通管制功能自动化并转移到飞机上。这为发展这里提出的理论提供了一个极好的环境，我们将利用这个丰富的应用领域进行示例和实验室实验。这个项目的教育部分是斯坦福大学控制理论的新课程。这些课程将从控制理论和计算机科学的角度结合系统理论。此外，将建立一个新的实验室，其目标是研究、建模和控制复杂系统。

This project is devoted to research and teaching in the area of hybrid systems theory. The goal is to develop a control theory for complex networks of systems, each with its own objectives, decision-making, and information processing capabilities. The systems have dynamics which are primarily continuous-state and continuous-time, yet the aggregate network is in general too complex to model with continuous tools alone. Control theory to date has achieved tremendous success in the analysis and synthesis of single control systems, as well as the development of control laws for groups of systems which are connected together by point-to-point links so that information is received and processed synchronously at each subsystem. Using continuous-state tools, simple hierarchical control schemes have also been designed and used with great success. The control of an aircraft is an excellent example: the complex system of hydraulic valves which controls the surfaces of today's commercial jets is transparent to the pilot, due to a control layer in between that bounds the operation of the hydraulics and displays this simplified abstraction of the complex system to the pilot.We are now interested in designing control laws for systems which are vastly more complex than continuous, synchronous control theory allows. Examples include air traffic control systems, highway systems, communication systems, the interconnected power grid, financial networks such as the stock market, and networks of distributed MEMS sensors. In each of these examples, the sensors and resources are distributed across the systems, yet there is a strong need for the systems to coordinate to perform some task or achieve some goal. The systems are also asynchronous, since correlated but different information is available to the controllers at different times. We propose to design hierarchical, hybrid control systems: ours will be a hierarchy in which discrete state models are used at the highest level to model the large mode spaces in the system, and in which the lowest level incorporates distributed sensing and control, asynchronous receipt and processing of information, and multiple objective functions. Our research will focus on:1. Design of the real-time hybrid interface: A key focus of this research is on the development of a robust and computationally efficient hybrid system interface between the discrete-state system abstraction and the continuous-state subsystems. This interface will provide a priori proofs that the system will function as desired, and automatic synthesis of both discrete and continuous control laws.2. Control of asynchronous systems: Hand-in-hand with autonomous interacting agents comes the issue of asynchrony. In classical control theory, sensor data is sampled and actuator commands are issued at a fixed rate. We propose to design an asynchronous control theory, in which sensors transmit time-stamped data packets whenever new information is available and controllers pick up packets relevant to their current function.3. Algorithms for automated air traffic systems: In future air transportation systems, it is proposed to automate and move much of the current air traffic control functionality on board the aircraft. This provides a superb environment in which to develop the theory proposed here, and we will draw on this rich application domain for examples and laboratory experiments.The education component of this project is a new curriculum in control theory at Stanford. The courses will combine systems theory from both control theoretic and computer science perspectives. In addition, a new Laboratory will be developed, with the goal of studying, modeling, and controlling complex systems.