基于门控循环单元的链路质量预测

链路质量预测可以为上层路由协议选择高质量的无线链路进行通信提供依据。采用手肘法优化的K-means++算法,在此基础上自适应划分链路质量等级；选择RSSI均值、LQI均值以及SNR均值作为链路质量参数；基于Catboost评估链路质量,采用网格搜索法优化CatBoost超参数;使用滑动时间窗口得到链路质量时间序列样本集,使用 GRU提取链路质量的时序信息,将时序信息输入支持向量回归预测得到下一时刻链路质量等级。在室内、走廊以及停车场三个场景下实验表明,与基于小波神经网络、循环神经网络和随机向量函数链等方法构建的链路质量预测模型相比,本文提出的方法有更高的预测准确率。

Link Quality Prediction Based on Gate Recurrent Unit

Link quality prediction can provide a basis for upper-layer routing protocols to select high-quality wireless links for communication. K-means++ algorithm optimized by elbow method is used to divide link quality level; Catboost is used to build link quality evaluation model, and grid search method is used to optimize evaluation model parameters to obtain link quality level; GRU (gate recurrent unit) is used to extract link quality time series information, and the timing information is input to support vector regression to predict the link quality level at the next moment. In the scenarios of indoor, corridor and parking lot, the experimental results show that compared with link quality prediction models based on wavelet neural network, random vector functional link, recurrent neural network and so on, the proposed method has better prediction accuracy.