Introduction - WilsonGuo/FastLivo_Replication GitHub Wiki

Background

The MaRS Laboratory at the University of Hong Kong is known for its prolific and high-quality research. FAST-LIVO is their work published in IROS 2022, which is a multi-sensor fusion solution combining LiDAR, IMU, and camera data. It was selected as one of the top ten best open-source SLAM algorithms of 2022.

One-Sentence Summary of Fast-LIVO Achievements:

"An extremely efficient, highly robust against various extreme degradation environments, pixel-level accuracy, the first LiDAR-visual-inertial odometry system applicable for UAV autonomous navigation, capable of real-time high-precision colored point cloud map reconstruction."

Fast-LIVO

A compact LiDAR-inertial-visual odometry framework built on two direct and tightly coupled odometry systems: the LIO subsystem and the VIO subsystem. These subsystems jointly estimate the system state by fusing their respective LiDAR or visual data with the IMU.

A direct and efficient VIO subsystem maximizes the reuse of the point cloud map built by the LIO subsystem. Specifically, points in the map are associated with previously observed image patches and aligned by projecting points onto new images to minimize direct photometric error (thus the complete system state). The reuse of LiDAR point clouds in the VIO subsystem avoids visual feature extraction, triangulation, and optimization, coupling the two sensors at the measurement level.

This system is open-source, capable of running in real-time on Intel or ARM processors, and supports multi-line rotating LiDAR and emerging solid-state LiDAR with entirely different scanning modes.

It is conceptually similar to the VINS-Fusion from Hong Kong University of Science and Technology (HKUST) (dual LIO and VIO systems).

Fast-LIVO GitHub Open Source Address

Replication Schedule

  1. Device Configuration and Assembly
  2. Environment Setup and Driver Debugging
  3. Hardware Synchronization
  4. Multi-Sensor Calibration (Camera and LiDAR)
  5. Running Fast-LIVO
  6. Performance Optimization