MRI Development: Heterogeneous, Autonomic Wireless Control Networks for Scalable Cyber-Physical Systems

提案编号：CNS 09-23518机构：丹佛大学(s): Voyles, Richard M.；科罗拉多州丹佛80208-0000拉胡尔曼格哈拉姆；Anaraki Siavash Pourkamali；马修·卢瑟福；Kimon p . Valavanis标题：MRI/Dev项目建议：该合作项目创建了一种由新型异构无线传感器-执行器-控制器平台组成的仪器，促进了网络网络物理系统（CPS）的广泛实验研究。一个关键目标是在平台类别上实现硬件和软件接口的标准化，这将支持时间和安全关键应用程序的协议。涉及四所大学（丹佛大学、圣母大学、宾夕法尼亚大学和阿灵顿大学），三种类型的网络化CPS研究平台在广泛的基于硬件和软件的运行时重新配置中开发。目标还包括在这些平台上开发标准化的硬件和软件接口，以便节点可以即插即用，在运行时参数化和程序化地发展，并保持控制和驱动连接对象的及时性和可靠性。宾夕法尼亚大学和丹佛大学现有的计算节点原型将被改进和协调，以提供一套可互操作的节点。这些节点将具有用于数据平面的双无线电和用于基于细粒度硬件的全局时间同步的无源模拟无线电，以增加确定性。嵌入式虚拟机（EVM）是一种功能强大的分布式运行时系统，其虚拟组件及其属性是跨节点边界维护的，它可以在给定物理网络时空变化的情况下，跨一组控制器维护一组功能不变量，如控制律和准功能不变量，如时效性约束、容错和安全标准。EVM软件允许跨不同硬件类别的紧密耦合通信和运行时控制。编程机制将一组物理传感器、执行器和控制器视为单个虚拟组件，并允许在运行时分配任务，因为无线系统的链路、节点和拓扑本质上是不可靠的。该系统有望降低硬件、软件和虚拟自主计算结构、异构传感器网络定时、同步和任务分配策略等可重构计算研究的障碍，并作为生物医学建模、人类监视和监测以及搜索和救援机器人应用的跳板。每个节点都将连接到一套带有相应传感器和执行器的模块化I/O设备。最近对未来无线传感器网络和应用的研究活动仅限于开环传感和监测，从而产生了主要基于事件的异步平台和系统软件。对于集成了一系列计算和通信能力的异构无线传感器网络的研究并不多。当网络与高速率传感器（如视频监控）、具有时效性和安全性约束的执行器（如实时控制）以及需要大量分布式网络内处理的网络（如视频分析和自主系统）集成时，调查人员必须超越低速率传感器和时间戳足够的应用平台。因此，集成计算和通信能力的异构无线传感器网络是必要的。更广泛的影响：该项目涉及四个机构，提供一系列可互操作的控制节点，以开发从MEMS/NEMS（微/纳米机电系统）规模到宏观规模的应用，开发用于搜索和救援、工业自动化、医疗设备和车辆控制的无线控制网络的构建模块。学生参与开发仪器。

Proposal #: CNS 09-23518 Institution: University of DenverPI(s): Voyles, Richard M.; Denver, CO 80208-0000 Mangharam, Rahul; Anaraki Siavash Pourkamali; Rutherford, Matthew J.; Valavanis, Kimon P.Title: MRI/Dev.: Heterogeneous, Autonomic Wireless Control Networks for Scalable Cyber-Physical SystemsProject Proposed:This collaborative project, creating an instrument consisting of a new class of heterogeneous wireless sensor-actuator-controller platforms, facilitates a wide range of experimental research on Networked CyberPhysical Systems (CPS). A key aim is to arrive at standardization for hardware and software interfaces over the platform categories that will support protocols for time- and safety-critical applications. Involving four universities (U Denver, Notre Dame U., U Penn, and UT-Arlington), three categories of Networked CPS research platforms are developed across a wide range of hardware- and software-based runtime re-configuration. The goals also include developing standardized hardware and software interfaces across these platforms so that nodes may be plug-n-play, evolve parametrically and programmatically at runtime, and maintain timeliness and reliability as connected objects for control and actuation. Existing computational node prototypes from Penn and U Denver will be refined and harmonized to provide a suite of interoperable nodes. These nodes will have dual radios for the data-plane and a passive analog radio for fine-grained hardware-based global time synchronization to add determination. An Embedded Virtual Machine (EVM), a powerful distributed runtime system where virtual components and their properties are maintained across node boundaries, is introduced to maintain a set of functional invariants, such as control law and para-functional invariants such as timeliness constraints, fault tolerance and safety standard across a set of controllers given the spatio-temporal changes in the physical network. The EVM software allows tightly coupled communication and runtime control across the different hardware categories. Programming mechanisms treat the set of physical sensors, actuators, and controllers as a single virtual component and allow tasks to be assigned at runtime since the links, nodes, and topology of wireless systems are inherently unreliable. The system is expected to lower the barriers for research into reconfigurable computing across hardware, software, and virtual autonomic computing structures, heterogeneous sensor network timing, synchronization and task allocation strategies, and also serve as a springboard to applications in biomedical modeling, human surveillance and monitoring, and search and rescue robotics. Each node will interface to a suite of modular I/O devices with attendant sensors and actuators. Recent research activity on future wireless sensor networks and applications has been limited to open-loop sensing and monitoring giving rise to predominantly event-based, asynchronous platforms and systems software. Not much research has been devoted to heterogeneous wireless sensor networks that integrate across a range of computational and communication capabilities. When networks are integrated with higher-rate sensors (e.g., video surveillance), actuators with timeliness and safety constraints (e.g., real-time control), and networks requiring significant distributed in-network processing (e.g., video analytics and autonomous systems), investigators have to go beyond the platforms for low-rate sensors and applications for which time-stamping is sufficient. Consequently, heterogeneous wireless sensor networks that integrate computational and communication capabilities are necessary.Broader Impacts: The project, involving four institutions, provides a range of interoperable control nodes to develop applications from the MEMS/NEMS (Micro/Nano ElectroMechanical Systems) scale to the macro scale, develops building blocks for wireless control networks with applications in search and rescue, industrial automation, medical devices and vehicular control. Students are involved in developing the instrumentation.