Collaborative Research: CSR-EHS: Obtaining Realistic Communication and Sensing In-situ Models for Wireless Embedded Systems

随着MEMS技术的不断进步，无线嵌入式系统被部署在各种环境中，从控制良好的实验室到动荡的海底。系统设计对产生的通信和感知模式进行了简化，有时甚至是不切实际的假设。虽然嵌入式设备通常在部署前单独进行微校准，但初步结果表明，部署环境是影响嵌入式设备通信和传感特性的主要因素。本研究旨在为现实环境下的大规模嵌入式系统开发广泛的建模方法和相关协议。该项目的主要目标在于开发三种新颖的建模方法，它们相互补充，覆盖了很大的成本效益设计空间。首先，基于不规则度和信号功率方差的概念，建立了一个新的无线电不规则度模型。第二种是使用自动捕获和重放过程，寻求对完全可重复的物理环境进行抽象的能力。第三个特点是一种以受控方式使用训练事件来产生非参数现实传感和通信模式的新方法。在资源有限的大型嵌入式系统中，如何协调原位建模精度与建模成本之间的矛盾是原位建模面临的一个关键挑战。本项目寻求从微观到宏观的模型开发，设计师可以根据精度需求选择合适的细节级别，以及从参数到非参数类型的模型，设计师可以根据可用资源选择合适的成本模型。这些模型将通过通用仿真系统提供，使嵌入式系统设计人员能够开发基于真实感的解决方案，并避免目前发现的由仿真开发的解决方案在现实世界中不起作用的普遍问题。反过来，预计这将通过节省开发时间和金钱并产生更高效、健壮、可预测和可控的系统来对嵌入式系统产生重大影响。研究结果将通过明尼苏达大学和弗吉尼亚大学的交通监控和管理测试平台进行评估。

With the continuing advancement of MEMS technology, wireless embedded systems are deployed in various kinds of environments from well-controlled laboratories to turbulent ocean floors. System design has made simplifying and sometimes unrealistic assumptions about the resulting communication and sensing patterns. Although embedded devices are normally micro-calibrated individually before deployment, preliminary results indicate that the deployment environment is a dominating factor in communication and sensing characteristics of embedded devices. This research aims at developing a wide spectrum of modeling methodologies and related protocols for large-scale embedded systems under realistic environments. The main objective of this project lies in developing three novel modeling approaches, which complement each other and cover a large cost-benefit design space. The first is to develop a new radio irregularity model based on concepts of degree of irregularity and variance of signaling power. The second seeks a capability for abstraction of a completely repeatable physical environment, using an automated capture-and-replay process. The third features a novel way to use training events in a controlled manner to produce non-parametric realistic sensing and communication patterns. A key challenge for in-situ modeling lies in reconciling the conflict between the in-situ modeling accuracy and the related cost to build and use these models in resource-limited large-scale embedded systems. This project seeks to develop the models from the micro to macro levels where designers can choose the appropriate level of detail based on the accuracy needs, and also the models from parametric to non-parametric types where designers can choose the models with proper costs based on the available resources. The models will be available via common simulation systems, enabling embedded systems designers to develop solutions based on realism and avoid an all-to-common problem found today where solutions developed by simulation don't work in the real world. This, in turn, is expected to have a major impact on embedded systems by saving development time and money and resulting in more efficient, robust, predictable and controllable systems. Research results will be evaluated using traffic-monitoring and management testbeds at the University of Minnesota and the University of Virginia.