数据驱动下的传感器网络定位方法研究

Sensor network localization is a prerequisite to ensure normal operation of a network. The performance of traditional localization approaches depends on the quality of communication signals and the accuracy of parameter estimation. But they are not aware of the location-related valid information that sensor data carries. In this project, we hope to fully excavate the correlation between network topology and node spatial distribution through data analysis and processing technology based on graph structure, so as to get rid of the dependence of localization approaches on influential factors, e.g., range accuracy and distribution of anchor nodes, and extend their application scope. The main studies to be carried out in this project include: (1) Sensor data analysis and preprocessing to improve data accuracy and realize multi-sensor information fusion; (2) Research on spatial-temporal distribution of nodes driven by sensor data, focusing on network topology learning; (3) Research on graph-structure-based distributed computing methods so as to improve the scalability of proposed localization approaches. This project is not limited to specific types of sensor data. Proposed localization approaches are also suitable for the localization and tracking of other intelligent devices (such as mobile phones, robots) with different types of environment sensing sensors.

传感器网络定位是保证网络正常运行的前提。传统定位方法大多依赖于节点所采集通信信号的质量与参数估计的精度，而未意识到传感器数据本身所携带的与定位相关的有效信息。本项目希望通过基于图结构的数据分析与处理技术，充分挖掘网络拓扑结构与节点空间分布之间的相关性，从而摆脱定位方法对于测距精度、锚节点分布等因素的依赖，扩展定位方法的适用范围。本项目拟开展的主要研究内容包含：（1）传感器数据分析与预处理，用以提升数据精度、实现多传感器信息融合；（2）数据驱动下的节点时空分布研究，重点考虑网络拓扑结构学习；（3）基于图结构的分布式计算方法研究，着力提升定位方法的可扩展性。本项目研究并不局限于特定类型的传感器数据，因此所提出的定法方法同样适用于其它具有多种类型环境感知传感器的智能设备（诸如手机、机器人）定位与追踪。

传感器网络定位是保证网络正常运行的前提。传统定位方法大多依赖于节点所采集通信信号的质量与参数估计的精度，而未意识到传感器数据本身所携带的与定位相关的有效信息。本项目希望通过基于图结构的数据分析与处理技术，充分挖掘网络拓扑结构与节点空间分布之间的相关性，从而摆脱定位方法对于测距精度、锚节点分布等因素的依赖，扩展定位方法的适用范围。本项目拟开展的主要研究内容包含：（1）传感器网络图结构学习与逼近算法；（2）传感器数据预处理与图滤波器设计；（3）基于图结构的传感器网络定位算法。