基于多粒度流形学习的无线传感器网络定位方法*

针对FastMDS-MAP定位算法存在对不规则无线传感器网络定位误差大，选取的框架节点不能很好的体现网络的拓扑结构实现不同粒层定位的问题，通过选择不同的筛选半径获得不同粒度的框架节点，结合绝对坐标变换加权策略提出了基于多粒度流形学习的无线传感器网络定位方法(MG-MDS)。仿真实验结果表明，不规则网络中MG-MDS算法定位精度比FastMDS-MAP算法有明显的提高；且定位误差随着网络节点粒度的变细而变小。

A Multi-granularity-Based Manifold Learning Method for Localization in Wireless Sensor Networks

Aimed at those problems that Fast MDS-MAP localization algorithm has the high location error in irregularly shaped wireless sensor networks and is unable to select different granular levels of the network to locate, this paper proposed a multi-granularity-based manifold learning method for localization in wireless sensor networks, abbreviated as MG-MDS. Different granular framework nodes can be obtained by selecting the different filter radius, and a new strategy is introduced for transforming relative coordinates to absolute coordinates. Experimental results show that the MG-MDS algorithm can get the higher locating accuracy than the Fast MDS-MAP algorithm in irregular wireless sensor networks, and the localization error will be smaller when the granularity size of the network becomes finer.