Leg-KILO 2.0

Robust Kinematic-IMU-Lidar Odometry

Leg-KILO 2.0 is a kinematic–inertial–LiDAR tightly‑coupled error‑state Kalman filter odometry system. Both the methodology and implementation differ from the original paper. The new version of Leg‑KILO is more efficient and lightweight. Key features include:

• Tight Sensor Fusion via ESKF
  All sensors (LiDAR, IMU, and optional leg kinematics) are fused in a single Error‑State Kalman Filter.

• Per-Point LiDAR Observations & IMU as Model Observation
  Each LiDAR point is treated as an independent observation, and the IMU is used as a model observation—inspired by Point‑LIO. This makes the system more robust during high dynamic motion.

• Voxel Map Management
  A voxel‑based map (based on FAST‑LIVO2) is used to organize and manage LiDAR map.

• High Throughput
  Thanks to the ESKF and voxel map structure, single‑frame processing runs in 5–20 ms .

• Extensive Validation
  Tested on both self‑collected and public datasets, and validated on Unitree Go1 and Go2 robots(with more datasets under continuous testing).

News

• 2024.07.20: The paper is accepted by RA-L 2024!
• 2024.07.31: The code is released.
• 2025.07.20: Leg-KILO 2.0 is released.

Prerequisites

Does not include any external optimization libraries; only requires common SLAM libraries such as Eigen and PCL.

Currently our code is tested on

• Ubuntu 18.04
• ROS melodic
• pcl 1.8
• eigen 3
• unitree_legged_msgs (has included in the project)
• glog
• yaml-cpp

sudo apt update && sudo apt install -y libpcl-dev libeigen3-dev libgoogle-glog-dev libyaml-cpp-dev

Build

cd ~/legkilo_ws/src
git clone https://github.com/ouguangjun/Leg-KILO.git
cd ..
catkin build  # catkin_make

Run

Leg-KILO Dataset

Download our dataset from link

source devel/setup.bash
roslaunch legkilo leg_fusion.launch
rosbag play xxxx.bag

Diter++ Dataset

Download Diter++ dataset from link

source devel/setup.bash
roslaunch legkilo diter.launch
rosbag play lawn_go2_lower_day.bag

NCLT Dataset

source devel/setup.bash
roslaunch legkilo nclt.launch
rosbag play xxxx.bag

Acknowledgments

Thanks for their excellent open source work:

• Point‑LIO
• FAST‑LIVO2
• FASTER-LIO
• SAD
• SVO
• A1-QP-MPC-Controller.

Citation

If you found this code/work to be useful in your own research, please considering citing the following information.

@ARTICLE{legkilo,
  author={Ou, Guangjun and Li, Dong and Li, Hanmin},
  journal={IEEE Robotics and Automation Letters}, 
  title={Leg-KILO: Robust Kinematic-Inertial-Lidar Odometry for Dynamic Legged Robots}, 
  year={2024},
  volume={9},
  number={10},
  pages={8194-8201},
  doi={10.1109/LRA.2024.3440730}}

Contact

If you have questions, make an issue or contact me at ouguangjun98@gmail.com

Star History