Sensors: Control-Theoretic Heterogeneous Design of Distributed Sensing-Network Algorithms: Exploiting Topological Structure and Overcoming Constraints

我们提出的研究是关于为具有通信/感知能力的代理网络设计分布式算法。我们的新算法将能够利用通信/传感拓扑的特点和设计范例中的异构性，以便在面对约束(如执行器饱和或通信速率约束)和随机或非随机拓扑变化时完成协调任务。因此，这些算法将非常适合于现代应用，如自动车辆协调、基于传感器网络的检测和空中交通流量管理等。我们提出的设计这些传感器网络算法的方法本质上是控制理论。我们将利用几个控制理论工具|最突出的是，使用分散控制理论中固定模式的移动和影响模型范例的工具|来深入了解分布式算法的性能和设计。特别是，我们将系统地刻画和设计一系列规范任务的算法，包括编队、同意和自划分任务，以及内部耦合网络(如空中交通系统或电力系统)的任务。对于每个规范任务，我们的目标将是确定在面对现实内容和拓扑变化时实现高性能所需的拓扑结构和设计异构性程度。潜在的控制理论工具将根据我们的算法设计需求进行扩展(例如，它们将被扩展到考虑到拓扑变化)，并将进行复杂任务的仿真和实验。该提议的智能优点是显而易见的：为了开发新的传感网络算法，解决了困难的分布式控制问题。我们研究的更广泛的影响是重大的：华盛顿州立大学已经设计了一门关于网络动力学和控制的新课程，目前正在这里授课，并将努力培养人们对传感网络及其控制的兴趣和传播信息。

Our proposed research is concerned with the design of distributed algorithms for networks of agentswith communication/sensing capabilities. Our novel algorithms will be capable of exploiting fea-tures of the communication/sensing topology and heterogeneities in the design paradigm in order tocomplete coordinated tasks in the face of constraints (such as actuator saturation or communicationrate constraints) and stochastic or nonrandom topological variations. Hence, the algorithms will bewell-suited for modern applications such as autonomous-vehicle coordination, sensor network-baseddetection, and air trafficc flow management, among others.Our proposed methodology for designing these sensing-network algorithms is control-theoreticin nature. We will draw on several control-theoretic tools|most prominently, tools that use thenotion of a fixed-mode in decentralized control theory and the influence model paradigm|to gaininsight into distributed algorithm performance and design. In particular, we will systematicallycharacterize and design algorithms for a sequence of canonical tasks, including formation, agree-ment, and self-partitioning tasks, and tasks for internally-coupled networks (such as Air TrafficSystems or electric power systems). For each canonical task, our goal will be to identify the topo-logical structure and degree of design heterogeneity needed to achieve high performance in theface of realistic contstraints and topological variation. The underlying control-theoretic tools willbe extended based on our algorithm-design needs (e.g., they will be extended to be account fortopological variation), and simulations and experiments of complex tasks will be performed.The intellectual merit of the proposal is evident: difficult distributed control problems areaddressed to in order to develop novel sensing-network algorithms. The broader impact of ourresearch is significant: a new course on network dynamics and control has been designed and iscurrently being taught here at Washington State University, and efforts will be made to fosterinterest and disseminate information about sensing networks and their control.