Paper Abstract and Keywords |
Presentation |
2021-05-14 10:25
[Invited Lecture]
Matrix Completion Based Missing RSS Sequence Recovery and Future Value Prediction Norisato Suga (ATR/TUS), Kazuto Yano (ATR), Julian Webber (ATR/Osaka Univ.), Yafei Hou (ATR/Okayama Univ.), Eiji Nii, Toshihide Higashimori, Yoshinori Suzuki (ATR) CQ2021-12 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
This paper proposes rank minimization and matrix factorization (MF) based interpolations 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 paper, we propose to apply an rank minimization and MF-based interpolation techniques to RSS interpolation in order to restore the true RSS variation pattern.
The simulation results show that the proposed interpolations can improve the prediction accuracy of the machine learning based channel prediction method. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
RSS prediction / rank minimization / matrix factorization / / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 121, no. 15, CQ2021-12, pp. 47-52, May 2021. |
Paper # |
CQ2021-12 |
Date of Issue |
2021-05-06 (CQ) |
ISSN |
Online edition: ISSN 2432-6380 |
Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
Download PDF |
CQ2021-12 |
|