CAREER: Design and Analysis of Performance-Critical Wireless Sensor Networks: A Fusion-Centric Approach

无线传感器网络(WSNs)具有以前所未有的时空尺度监控物理世界的潜力。最近，无线传感器网络被部署用于许多新兴的性能关键型应用，例如监控重要的基础设施(电网和桥梁)和检测自然灾害(火山和地震)。然而，无线传感器网络与物理世界深度结合，经常受到环境噪声、物理现象的动态变化和网络部署不准确等不确定性造成的显著性能变化。该项目开发了一种原则性的网络设计和分析方法来保证无线传感器网络的性能。与现有的基于启发式的解决方案不同，该方法采用数据融合这一先进的信息处理方案，使传感器能够高效协作，提供可预测的网络性能。该项目有四个目标：1)一个基于已建立的数据融合模型的时空感知性能分析框架。该分析捕获了由事件检测定义的时空覆盖与虚警概率、融合模型、网络密度以及包括噪声和部署不准确性在内的物理不确定性之间的基本关系。2)数据融合方案，它利用移动传感器来重新配置网络的能力，以响应物理动态。3)统一的融合和通信架构抽象，允许开发人员使用组级原语实现和优化网络协议。4)一种模型驱动的媒体访问控制协议，能够在协同数据处理中获得可预测的吞吐量和延迟。该项目将对大量对感知和通信性能有严格要求的关键应用产生影响。该项目的成果将纳入外展活动、课程开发和学生辅导。

Wireless Sensor Networks (WSNs) have the potential to monitor the physical world at an unprecedented spatio-temporal scale. Recently, WSNs have been deployed for many emerging performance-critical applications such as monitoring important infrastructures (power grid and bridges) and detecting natural hazards (volcanoes and earthquakes). However, deeply integrated with the physical world, WSNs often suffer from significant performance variations caused by uncertainties including environmental noise, dynamics of physical phenomena, and network deployment inaccuracy. This project develops a principled network design and analysis approach to performance assurance of WSNs. In contrast to existing heuristics-based solutions, this approach adopts data fusion, an advanced information processing scheme, to enable sensors to efficiently collaborate in delivering predictable network performance. This project has four aims: 1) A framework for analyzing spatio-temporal sensing performance based on established data fusion models. The analysis captures fundamental relationship between spatio-temporal coverage defined by event detection and false alarm probabilities, fusion models, network density, and physical uncertainties including noise and deployment inaccuracy. 2) Data fusion schemes that exploit mobile sensors to reconfigure the capability of a network in response to physical dynamics. 3) A unified fusion and communication architecture abstraction that allows developers to implement and optimize network protocols using group-level primitives. 4) A model-driven medium access control protocol that can achieve predictable throughput and delay in collaborative data processing.This project will have impact in numerous critical applications that require stringent sensing and communication performance. The results of this project will be integrated into outreach activities, curriculum development, and student mentoring.