GIVL-SLAM: A Robust and High-Precision SLAM System by Tightly Coupled GNSS RTK, Inertial, Vision, and LiDAR
Xuanbin Wang, Xingxing Li, Hui Yu, Hanyu Chang, Yuxuan Zhou, Shengyu Li
Abstract
In this article, we present GIVL-SLAM, a factor graph optimization-based framework that tightly fuses double-differenced pseudorange and carrier phase observations of the GNSS with inertial, visual, and LiDAR information for high-level simultaneous localization and mapping (SLAM) performance in large-scale environments. A sliding-window-based factor graph estimator is designed to explore the potential of heterogeneous observations from multiple sensors for achieving robust and high-accuracy state estimation. The integer ambiguity resolution of GNSS carrier phase observation is also considered in our method to leverage the high-precision characteristic of the carrier phase. We extensively evaluated the proposed method in real-world experiments including both GNSS-challenged environment tests and urban night environment tests. The results demonstrate that the proposed GIVL-SLAM significantly improves the global drift-free ability of the visual-inertial-LiDAR system in large-scale conditions and achieves continuous centimeter to decimeter-level localization performance in GNSS harsh-signal conditions. The maximum improvement of the 3-D location availability (<1 m) in GNSS severely degraded situations is more than 60% compared with the existing loosely coupled GNSS/SLAM fusion methods.