書誌事項
- タイトル別名
-
- A Proposal and Evaluation of Positioning Method Based on Wi-Fi RTT and Machine-learning
この論文をさがす
説明
屋外の測位では一般的にGPSが利用されているが,屋内では衛星からの電波が届かないためGPSによる測位は難しい.そのため様々な屋内測位手法が研究されている.無線LANを用いた測位では,無線LANの電波強度をあらかじめ測定し,ベイズ推定で測位する試みがあるが,電波の特性上精度に課題が残る.一方で近年,無線LANのラウンドトリップタイムを利用し,1~2m以内の精度で距離推定を可能としたIEEE802.11mcが標準化され,スマートフォンでもAndroid 9以降からIEEE802.11mcが利用可能になってきた.本研究では,電波強度だけではなくアクセスポイントと端末の距離のデータを学習させ,位置を推定する手法を提案し評価した.実験では3機の無線LANアクセスポイントを用意し,位置推定の評価方法として,ベイズ推定,サポートベクタマシン,k-近傍法を用い比較した.実験の結果,学習に電波強度と距離のデータを併用した場合は,それぞれを単独で学習に用いた場合の精度を上回ることを示した.
GPS is generally used for outdoor positioning, but indoor positioning by GPS is difficult because indoors do not receive radio waves from satellites. Therefore, various indoor positioning methods have been studied. In the position estimation using the wireless LAN, there is an attempt to use the Bayesian estimation by using the radio strength measured in advance, but there remains a problem in the accuracy of the radio wave. On the other hand, in recent years, IEEE802.11mc, which enables distance estimation with accuracy within 1-2m using round-trip time of wireless LAN have been standardized and available on Android 9. In this research, we proposed and evaluated a positioning method based on the machine-learning using not only the radio wave intensity but also the data of the distance between the access point and the terminal. In the experiment, three wireless LAN access points were prepared, and Bayesian estimation, support vector machine, and k-nearest neighbor method were used as the evaluation method for position estimation and comparison was performed. As a result of the experiment, it was shown that the combined use of the radio wave intensity and the distance data for learning exceeds the accuracy when each is used alone for learning.
収録刊行物
-
- 情報処理学会論文誌
-
情報処理学会論文誌 62 (2), 465-474, 2021-02-15
dummy publisher
- Tweet
詳細情報 詳細情報について
-
- CRID
- 1390572174709975168
-
- NII論文ID
- 170000184378
-
- NII書誌ID
- AN00116647
-
- ISSN
- 18827764
-
- Web Site
- http://id.nii.ac.jp/1001/00209319/
-
- 本文言語コード
- ja
-
- 資料種別
- journal article
-
- データソース種別
-
- JaLC
- IRDB
- CiNii Articles
- KAKEN