MOdel-based approaches for enhancing wireless sensor network usability, programmability, reliability and efficiency

虽然无线传感器网络（wsn）作为一种监测工具提供了巨大的潜力，但它们目前的编程和使用非常复杂。基于许多类型的传感器数据在它们变得有用之前需要合并到模型中的观察，我建议通过将它们与物理和决策上下文的显式模型集成来降低wsn的复杂性并提高功率经济性和可靠性。基于模型的物理环境规范和WSN对物理环境的响应将大大降低应用程序实现的复杂性。将显式模型集成到WSN应用程序开发中还允许模型编译器使用所需的信息自动将与模型相关的计算分布到WSN的节点上。这种基于模型的wsn不是将所有传感器数据转发到网络外进行外部处理，而是负责以分布式方式内部跟踪和报告物理环境状态的“最佳猜测”。网络内模型的执行将通过减少通信来改善电力经济，并通过减少对外部中央系统通信的依赖来增加监控应用程序的鲁棒性。我们计划为流行的应用程序类提供面向用户的数据库接口、编程语言和相关的编译器，以及支持以模型为中心的应用程序开发所需的支持库。TinyDB WSN数据库系统将进行调整，允许用户使用从模型域提取的维度和自然概念（如时间和可能性）对模型状态和输出进行临时或定期的查询。在这个项目的后期阶段，我们将扩展所涉及的模型并引入决策上下文的模型。通过整合对决策需要的数据含义的理解，这种集成将允许改善能源和传感器数据之间的权衡。我们希望这项工作能够大大降低在设施、工业过程、农场和民用结构中部署基于无线传感器的监测的成本效益障碍，从而造福社会。通过提高wsn重要子类的可用性、可编程性和功率效率，这些项目将促进有效监测应用的发展。在涉及监测设施、工业过程、农场和民用结构的应用方面，作出贡献的机会特别大。

While wireless sensor networks (WSNs) offer great potential for use as a monitoring tool, they are currently dauntingly complex to program and use.  Building on the observation that many types of sensor data require incorporation into a model before they can become useful, I propose to reduce complexity and improve power economy and reliability in WSNs by integrating them with explicit models of physical and decision context.  Model-based specification of the physical environment and a WSN's response to it will greatly reduce application implementation complexity.  Incorporating explicit models into WSN application development also permits a model compiler with the information required to automatically distribute model-related computation to nodes across the WSN.  Rather than relaying all sensor data out of the network for external processing, such model-based WSNs will instead be responsible for internally tracking and reporting on a "best guess" of the state of the physical environment in a distributed fashion.  In-network model execution will improve power economy by reducing communication, and increase monitoring applications' robustness by lessening reliance on communication to an external central system.       We plan to provide user-oriented database interfaces, programming languages and associated compilers for popular classes of applications, and supporting libraries needed to support model-centric applications development.  The TinyDB WSN database system will be adapted to allow ad hoc or periodic user queries regarding model state and output using dimensions drawn from the model domain and natural concepts such as time and likelihoods.  In the later stages of this project, we will expand the models involved and introduce models of decision context.  By incorporating an understanding of the implications of data for decision-making needs, this integration will permit improved tradeoffs between energy and sensor data.      We expect this work to benefit society by substantially lowering the barriers to cost-effective deployment of wireless sensor-based monitoring in facilities, industrial processes, farms, and civil structures. By enhancing the usability, programmability, and power efficiency of important subclasses of WSNs, these projects should facilitate the development of effective monitoring applications.  Opportunities for contribution are particularly strong in applications involving the monitoring of facilities, industrial processes, farms, and civil structures.