hku-mars/FAST_LIO
A computationally efficient and robust LiDAR-inertial odometry (LIO) package
What it solves
FAST-LIO provides a computationally efficient and robust way for robots to track their position and build maps using a combination of LiDAR (laser scanning) and IMU (inertial measurement) sensors. It is specifically designed to maintain high accuracy and stability in challenging scenarios, such as fast-moving environments, cluttered spaces, or situations where sensor data might otherwise degrade.
How it works
The system uses a "tightly-coupled" approach, meaning it fuses raw LiDAR feature points and IMU data together within an iterated extended Kalman filter. This filter continuously updates the robot's state estimate. The system includes an incremental mapping feature using a specialized data structure (ikd-Tree) to perform fast searches, allowing it to process LiDAR data at high rates without requiring complex feature extraction.
Who it’s for
This project is intended for robotics researchers and developers working on navigation, mapping, and autonomous systems. It is compatible with various LiDAR types, including spinning (e.g., Velodyne, Ouster) and solid-state sensors (e.g., Livox), and can run on resource-constrained hardware like ARM-based platforms (e.g., Raspberry Pi, NVIDIA TX2).
Highlights
- High-speed incremental mapping capable of exceeding 100Hz LiDAR rates.
- Direct odometry on raw LiDAR points, removing the need for manual feature extraction.
- Robust performance in fast-motion or noisy environments using a tightly-coupled Kalman filter.
- Broad hardware support, including various LiDAR models and ARM-based embedded computers.
Related
- Project
- Project
- Project
- Project