Matrix Factorization-Based RSS Interpolation for Radio Environment Prediction

Norisato Suga, Kazuto Yano, Julian Webber, Yafei Hou, Eiji Nii, Toshihide Higashimori, Yoshinori Suzuki

Research output: Contribution to journalArticlepeer-review

Abstract

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.

Original languageEnglish
JournalIEEE Wireless Communications Letters
DOIs
Publication statusAccepted/In press - 2021

Keywords

  • Artificial neural networks
  • Channel prediction
  • Indexes
  • Interpolation
  • matrix factorization
  • Minimization
  • Production facilities
  • received signal strength interpolation.
  • Robots
  • Wireless communication

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Electrical and Electronic Engineering

Fingerprint Dive into the research topics of 'Matrix Factorization-Based RSS Interpolation for Radio Environment Prediction'. Together they form a unique fingerprint.

Cite this