GNSS不感屋外環境における無人地上車両のための堅牢な自己位置推定ソリューション
A Robust Localization Solution for an Uncrewed Ground Vehicle in Unstructured Outdoor GNSS-Denied Environments
GNSSが使えない非構造屋外環境で走行する無人地上車両向けに、地形参照航法(TRN)と車輪オドメトリ・IMUをEKFで統合し、ループ閉じ込み不要で長距離・片道ミッションでも高精度を維持する自己位置推定システムTRAELSを提案した。
著者: W. Jacob Wagner, Isaac Blankenau, Maribel DeLaTorre, Amartya Purushottam, Ahmet Soylemezoglu
分類: cs.RO
原文アブストラクト
This work addresses the challenge of developing a localization system for an uncrewed ground vehicle (UGV) operating autonomously in unstructured outdoor Global Navigation Satellite System (GNSS)-denied environments. The goal is to enable accurate mapping and long-range navigation with practical applications in domains such as autonomous construction, military engineering missions, and exploration of non-Earth planets. The proposed system - Terrain-Referenced Assured Engineer Localization System (TRAELS) - integrates pose estimates produced by two complementary terrain referenced navigation (TRN) methods with wheel odometry and inertial measurement unit (IMU) measurements using an Extended Kalman Filter (EKF). Unlike simultaneous localization and mapping (SLAM) systems that require loop closures, the described approach maintains accuracy over long distances and one-way missions without the need to revisit previous positions. Evaluation of TRAELS is performed across a range of environments. In regions where a combination of distinctive geometric and ground surface features are present, the developed TRN methods are leveraged by TRAELS to consistently achieve an absolute trajectory error of less than 3.0 m. The approach is also shown to be capable of recovering from large accumulated drift when traversing feature-sparse areas, which is essential in ensuring robust performance of the system across a wide variety of challenging GNSS-denied environments. Overall, the effectiveness of the system in providing precise localization and mapping capabilities in challenging GNSS-denied environments is demonstrated and an analysis is performed leading to insights for improving TRN approaches for UGVs.
