LLIO: Lidar-Kinematic-Inertial Odometry with Ground Contact Constraints for Legged Robots
Chengjie Gu, Zhongqu Xie, Beichen Xiang, Shichao Zhou, Lingkun Chen, Binbin Ci, Yulin Wang
AI summary
Problem
High-frequency vibrations during legged robot locomotion degrade IMU and LiDAR data, causing rapid localization drift. Existing kinematic-inertial methods lack external constraints, while sensor-heavy approaches impose excessive computational overhead that compromises real-time performance.
Approach
LLIO fuses LiDAR, IMU, and joint encoder data using an iterated Kalman filter, augmented by a sliding-window ground contact constraint module to detect and correct attitude drift. A keyframe-based factor graph optimization then minimizes global cumulative drift.
Key results
- Filter-based LiDAR-kinematic-inertial odometry with factor graph back-end optimization
- Ground contact constraint module effectively compensates attitude drift on flat, rough, and sloped terrain
- Lower local drift and superior computational efficiency compared to existing LiDAR-kinematic-inertial methods
- Public release of the LLIO dataset for community benchmarking
Why it matters
Enables robust, real-time localization for legged robots in GPS-denied or complex environments without heavy computational overhead.
Abstract
This letter presents a robust multi-sensor fusion framework for state estimation in legged robots (LLIO) based on an iterated extended Kalman filter. To address the limitations of IMU priori estimation, which often leads to legged robot localization errors or failures, our method integrates the contact constraints of the robot’s leg kinematics with the ground. By introducing a sliding window-based ground contact constraint module, we effectively combine the contact state of the legged robot’s foot with ground features, enhancing the constraints in complex environments and reduce localization drift. Additionally, factor graph optimization minimizes global cumulative drift. The proposed method has been extensively evaluated through numerous experiments and relevant public datasets. The results demonstrate that our approach significantly reduces local drift and better computational efficiency.