LIGO: A Tightly Coupled LiDAR-Inertial-GNSS Odometry Based on a Hierarchy Fusion Framework for Global Localization With Real-Time Mapping
Dongjiao He, Haotian Li, Jie Yin
Abstract
This article introduces a method for tightly fusing sensors with diverse characteristics to maximize their complementary properties, thereby surpassing the performance of individual components. Specifically, we propose a tightly coupled light detection and ranging (LiDAR)-inertial-global navigation satellite system (GNSS) odometry (LIGO) system, which synthesizes the advantages of LiDAR, inertial measurement unit (IMU), and GNSS. Integrating LiDAR with IMU demonstrates remarkable precision and robustness in high-dynamics and high-speed motions. However, LiDAR-Inertial systems encounter limitations in feature-scarce environments or during large-scale movements. GNSS integration overcomes these challenges by providing global and absolute measurements. LIGO employs an innovative hierarchical fusion approach with both front-end and back-end components to achieve synergistic performance. The front-end of LIGO utilizes a tightly coupled, extended Kalman filter (EKF)-based LiDAR-Inertial system for high-bandwidth localization and real-time mapping within a local-world frame. The back-end tightly integrates the filtered LiDAR-Inertial factors from the front-end with GNSS observations in an extensive factor graph, being more robust to outliers and noises in GNSS observations and producing optimized globally referenced state estimates. These optimized back-end results are then fed back to the front-end through the EKF to ensure a drift-free trajectory, particularly in degenerate and large-scale scenarios. Real-world experiments validate the effectiveness of LIGO, especially when applied to aerial vehicles with outlier-prone GNSS data, demonstrating its resilience to signal losses and data quality fluctuations. LIGO outperforms comparable systems, offering enhanced accuracy and reliability across varying conditions.