This letter proposes matrix factorization (MF) based interpolation of received signal strength (RSS) from a transmitter mounted on moving robot in factory environment. For realizing the reliable wireless communication, machine learning based channel prediction methods have been intensively studied in the past decade. However, some traffic models will make the observation of RSS sequence be intermittent, and the missing values must be interpolated before input to the predictor. Classical interpolation such as linear interpolation cannot appropriately estimate the missing values because the result of the interpolation depends on the observation time. In this letter, we propose to apply an MF-based interpolation technique to RSS interpolation in order to restore the true RSS variation pattern. Moreover, the basic MF-based interpolation is improved by introducing a smoothing term in an objective function to represent the smooth variation of the RSS sequence. The simulation results show that the MF-based interpolation can improve the prediction accuracy of the machine learning based channel prediction method.

中文翻译：

---

基于矩阵分解的 RSS 插值用于无线电环境预测


这封信提出了基于工厂环境中移动机器人上安装的发射器接收信号强度 (RSS) 的基于矩阵分解 (MF) 的插值。为了实现可靠的无线通信，基于机器学习的信道预测方法在过去十年中得到了深入研究。然而，某些流量模型会使RSS序列的观察变得间歇性，并且必须在输入预测器之前对缺失值进行插值。线性插值等经典插值无法正确估计缺失值，因为插值的结果取决于观测时间。在这封信中，我们建议将基于 MF 的插值技术应用于 RSS 插值，以恢复真实的 RSS 变化模式。此外，通过在目标函数中引入平滑项来表示RSS序列的平滑变化，改进了基于MF的基本插值。仿真结果表明，基于MF的插值可以提高基于机器学习的信道预测方法的预测精度。