| 講演抄録/キーワード |
| 講演名 |
2025-07-01 09:00
点群の確率分布を考慮したLiDAR SLAM実装の試み ○内田 樹・岩切宗利(防衛大) |
| 抄録 |
(和) |
SLAM(Simultaneous Localization and Mapping)は,ロボットが未知環境を自律的に移動するために不可欠な技術である.中でも,LiDAR SLAM(Light Detection and Ranging)を用いたSLAM手法は,特定の環境下で高精度な自己位置推定および地図作成を実現できる手法として注目を集めている.これまでに,自己位置および姿勢の6成分の時間変化に基づくSLAM評価法を提案し,既存手法の課題を明確化してきた.本報告では,LiDAR点群の確率分布を考慮することで,従来手法が直面する課題を克服し得る新たなSLAM手法を提案する.特に,廊下や階段といった空間特有のSLAM退化を低減することを目的としている.提案手法の実証実験では,SubT-MRSを用い,従来のFast-LIOと,確率分布を考慮した点群を用いたFast-LIOとのSLAM精度を評価した.その結果,一部の評価指標について,提案手法は従来手法よりもSLAM退化を低減できることがわかった |
| (英) |
Simultaneous Localization and Mapping (SLAM) is an essential technology that enables robots to autonomously navigate unknown environments. Among the various SLAM approaches, those based on LiDAR (Light Detection and Ranging) have attracted significant attention for their ability to achieve high-precision self-localization and mapping in specific environments. In our previous work, we proposed a SLAM evaluation method that analyzes the temporal variation of six degrees of freedom in position and orientation, thereby revealing the limitations of existing SLAM methods. In this study, we present a novel SLAM method that addresses these limitations by incorporating the probabilistic distribution of LiDAR point clouds. The method is particularly designed to reduce SLAM degradation in environments, such as corridors and staircases. To validate the proposed approach, we conducted experiments using the SubT-MRS dataset, comparing the SLAM accuracy of the conventional Fast-LIO with that of an extended Fast-LIO that considers the probabilistic nature of point clouds. Experimental results indicate that, under certain evaluation metrics, the proposed method more effectively reduces SLAM degradation compared to the conventional approach. |
| キーワード |
(和) |
3次元点群 / 特徴抽出 / 自己位置推定 / 環境地図作成 / 確率分布 / / / |
| (英) |
Point-cloud / Feature Extraction / Localization / Mapping / Probability Distribution / / / |
| 文献情報 |
映情学技報, vol. 49, pp. 35-38, 2025年6月. |
| 資料番号 |
|
| 発行日 |
2025-06-23 (AIT) |
| ISSN |
Online edition: ISSN 2424-1970 |
| PDFダウンロード |
|