基于机器学习的5G室内定位方法

5G Indoor Positioning Method Based on Machine Learning

提出了一种适用于5G数字化室内分布式系统的通用室内定位方法。该方法首先采集数字化室分天线与终端(UE)之间的信道状态信息(Channel State Information,CSI),其次在时域和频域对CSI进行预处理,得到有效的CSI幅值,然后为减少CSI的重复采样和计算复杂度,构建灰色预测模型GM(1.1)对采集到的CSI样本值进行预测,生成新的CSI序列,最后采用机器学习算法构建室内信号传播模型,即CSI与距离之间的非线性关系,并结合多个天线完成定位。

It proposes a general indoor positioning method suitable for 5 G digital indoor division.This method first collects the Channel State Information(CSI)between the digital indoor antenna and the terminal(UE),and then preprocesses the CSI in the time domain and frequency domain to obtain an effective CSI amplitude,and then in order to reduce repeated sampling and computational complexity of CSI,the gray prediction model GM(1.1)is constructed to predict the collected CSI sample values,and a new CSI sequence is generated.Finally,the indoor signal propagation model is constructed by using machine learning algorithm,which describes the non-linear relationship between CSI and distance,and combined with multiple antennas,the positioning is completed.