Control of Automata-Switched Distributed Systems

该项目旨在开发使用分布式体系结构的动力系统控制的计算算法和理论，在分布式体系结构中，存在连续和逻辑动态的组合。这种情况是物理进程、通信进程和计算机进程耦合的结果。这项工作有非常广泛的相关应用，在这个项目中，我们特别强调它在联网机器人中的应用。所追求的具体方法和技术主要来自半定规划、图论、团队理论和鲁棒控制方法(例如，算子理论、二次成本函数和哈密尔顿-雅各比不等式，IQCs)，提供了使用分布式体系结构来处理和考虑基于计算机的规则、通信延迟和分组丢失、不确定环境中的障碍、轨迹跟踪和非线性动力学的统一框架。建议的研究的详细实施也将在PI的先进的分布式机器人试验台上进行，以提供明确的演示和使用该研究的设计示例。试验台的实施还将用于高中外展和本科生研究。该程序的主要目标是提供一种系统的方法，以凸优化和半定规划为基础，对同时具有离散和连续动态的系统进行分布式控制。它的出发点涉及国际和平研究所研究小组最近取得的两项独立成功。第一个涉及集中控制自动机切换系统的综合方法，自动机切换系统是由有限自动机驱动切换的切换系统。形成该程序基础的第二个最新成果是对一般图形拓扑进行分布式控制的工作。以上两个部分的结合为这个项目提供了起点，计划议程将处理理论、计算和实施。在这个项目中，有专门计划的机会，让本科生通过机电一体化项目，以及嵌入式系统软件和基于网络的软件的设计，参与研究。(2)本项目的研究目标是由计算机自动控制的物理系统，可能是通过互联网或使用无线技术。这类系统包括自主地面机器人，例如可用于并行分配系统、清理危险材料或大型建筑物的协调组装；以及无人驾驶飞行器，其具体应用可用于救灾期间的搜索和救援，或分布式检测。拥有创建机器人团队的能力，这些团队可以以类似于人类使用的方式进行协调，但可以信任以可预测的、唯一有益的方式执行，这是该项目的主要目标。虽然这一目标很重要，但实现这一业务能力仍然是一个巨大的障碍。关键问题是，尽管目前有可能设计和物理构建可以在没有人类帮助的情况下运行的高性能系统，但我们无法在任何程度上自信地预测它们在与环境自主交互时的表现，因为此类系统的控制算法必然非常复杂。该计划的主要目标和产品将是创造工具和算法，这些工具和算法可用于直接设计机器人，以及更广泛地说，具有安全保证的自主系统。

This project aims at developing computational algorithms and theory for control of dynamical systems using distributed architectures, where a combination of continuous and logical dynamics are present. This situation results from the coupling of physical, communications and computer processes. The work has a very broad set of associated applications, and in this project we particularly emphasize its use in networked robotics. The specific approach and techniques pursued stem mainly from semidefinite programming, graph theory, team theory, and robust control methods (e.g., operator theory, quadratic cost functions and Hamilton-Jacobi inequalities, IQCs), providing a unified framework in which to deal with and consider computer-based rules, communication latency and packet loss, obstacles in an uncertain environment, trajectory tracking, and nonlinear dynamics, using a distributed architecture. Detailed implementations of the proposed research will also be carried out on the PI's advanced distributed robotic testbed to provide explicit demonstrations and design examples using the research. The testbed implementations will also be used for high school outreach and undergraduate research. The major goal of the program is to provide a systematic approach to distributed control of systems that have both discrete and continuous dynamics, using convex optimization and semidefinite programming as the basis. Its starting point involves two separate recent successes in the PI's research group. The first involves synthesis methods for centralized control of automata-switched systems, which are switched systems where the switching is driven by a finite automaton. The second recent achievement that will form a basis for this program is work on distributed control over general graph topologies. The combination of the above two components provide the launching point for this project, and the program agenda will tackle theory, computation, and implementation. In this project there are specifically planned opportunities to involve undergraduate in research through mechatronics projects, and design of both embedded systems software and web-based software.(2) The research in this program targets physical systems that are to be automatically controlled by computers, possibly over the Internet or using wireless technology. Such systems include autonomous ground robots, which could for instance be used in parallel distribution systems, cleanup of hazardous materials, or coordinated assembly of large-scale structures; and unmanned aerial vehicles where specific applications could for instance be search-and-rescue during disaster relief, or distributed detection. Having the ability to create robotic teams that can coordinate in ways similar to those that humans use, but that can be trusted to perform in predictable, and only-beneficial, ways is the major goal of this program. Although this goal is an important one there remains a formidable barrier to achieving this operational capability. The key problem is that although it is currently possible to engineer and physically construct high-performance systems that can run without human assistance, we cannot predict with any degree of confidence how they will behave when interacting autonomously with their environment, because the control algorithms for such systems are necessarily very complex. The major goal and product of this program will be the creation of tools and algorithms that can be used to directly design robots and more generally autonomous systems that have safety guarantees.