基于机器学习的蜂窝网络节点定位算法研究

Research of the Node Localization Algorithm Based on Machine Learning for Cellular Networks

蜂窝网络希望能在广泛的应用领域内实现复杂的大范围监测和追踪任务,而移动台节点定位是相关应用的基础。本文在对现有无线网络定位技术研究的基础之上,有针对性地分析当前几种机器学习经典算法,提出了一种基于支持向量机树型多分类的蜂窝通信系统节点定位算法,充当分布式定位的全局坐标算法。通过对算法原理的分析以及实验结果对比,证明了基于机器学习的定位算法在定位效果方面解决了困扰基于信号参数的定位技术的边界问题与集中洞问题,在定位的平均误差、标准偏差和分布式定位正确率以及实现代价几个方面的总体性能均优于基于信号参数的定位技术与GPSone定位技术。

Cellular communication systems aim at achieving complex large-scale monitoring and tracing applications in wider fields, which is based on mobile station nodes localization. By studying the existing node localization technologies, this paper analyses the current several classical machine learning algorithms purposefully, and proposes a cellular communication system node localization algorithm based on machine learning, using it as a centralized coordinate algorithm of distributed node localization. Through simulation and theoretical analysis, it proves that the node localization algorithm in cellular communication systems based on machine learning can resolve the border problem and the coverage hole problem in the tra- ditional algorithms based on signal parameters, and its overall function is better than the traditional algorithms based on sig- nal parameters and GPS-one in terms of average error, standard deviation and the accuracy rate of distributed localization as well as the cost superior to the traditional location algorithm based on signal parameters.