Cognition and State Modeling

FAST-LIVO2

FAST-LIVO2 is a fast direct LiDAR-inertial-visual odometry system for real-time localization, mapping, and 3D reconstruction in degraded environments.

Tool Introduction

Core parameters, trigger timing, and visual before/after demo references.

Short Explanation

Play synchronized LiDAR, IMU, and camera data through FAST-LIVO2 and it estimates robot pose while building a local map in real time.

InputLiDAR + IMU + camera stream

OutputPose trajectory, local map, reconstruction

Trigger TimingTriggered when the ROS launch file receives synchronized sensor streams.

RuntimeROS / C++ / catkin / PCL / OpenCV

BeforeLiDAR + IMU + camera stream

Prepare the scene, image, video, sensor stream, prompt, or configuration expected by the original project.

AfterPose trajectory, local map, reconstruction

Read the produced visualization, prediction, map, trajectory, mask, grasp pose, or other documented artifact.

Preset Example

A quick-run style example for the documentation page.

Inputtools/fast-livo2/datasets/YOUR_DOWNLOADED.bag

Promptmapping_avia.launch with calibrated sensor topics

ExpectedA pose trajectory, local map, reconstruction outputs, and ROS visualization streams.

Parameters And Output

Readable controls and the meaning of each returned artifact.

Parameter Explanation

launch_fileselectmapping_avia.launch

Selects the sensor configuration and topic wiring for a specific LiDAR/camera setup.

configpathtools/fast-livo2/config/avia.yaml

Contains calibration, filter, mapping, and topic parameters.

rosbagfile

Recorded LiDAR, IMU, and image stream to replay.

save_maptogglefalse

Controls whether dense map artifacts are written after the run.

Output Explanation

trajectory

Estimated robot pose over time, usually evaluated with APE/RMSE.

map

LiDAR/visual reconstruction used for localization and inspection.

runtime

Per-frame processing time split across LiDAR and image updates.

How To Use

Official resources, deployment steps, academic context, citation, and source-reported benchmark numbers.

Resources

GitHubhttps://github.com/hku-mars/FAST-LIVO2 Code Downloadhttps://github.com/hku-mars/FAST-LIVO2/archive/refs/heads/main.zip Paperhttps://arxiv.org/abs/2408.14035 Launch Fileshttps://github.com/hku-mars/FAST-LIVO2/tree/main/launch FAST-LIVO2 Dataset Noteshttps://github.com/hku-mars/FAST-LIVO2#run-our-dataset Videohttps://youtu.be/6dF2DzgbtlY Datasethttps://github.com/xuankuzcr/Global-LVBA

Deployment Notes

Build the ROS catkin workspace after installing PCL, OpenCV, Ceres, and the dependencies listed by HKU-MARS.
Download a compatible rosbag and confirm camera-LiDAR-IMU calibration paths in the YAML config.
Run the matching launch file, then play the rosbag with simulated time if required.
Export trajectories, maps, and logs under tools/fast-livo2/runs/ for comparison.

Relative Path Example

# Relative-path local entry for the FAST-LIVO2 tool folder
cd tools/fast-livo2/catkin_ws
catkin_make
source devel/setup.bash

roslaunch fast_livo mapping_avia.launch
rosbag play tools/fast-livo2/datasets/YOUR_DOWNLOADED.bag

# Alternative launch/config files:
# tools/fast-livo2/launch/mapping_hesaixt32_hilti22.launch
# tools/fast-livo2/launch/mapping_ouster_ntu.launch
# tools/fast-livo2/config/avia.yaml
# tools/fast-livo2/config/NTU_VIRAL.yaml

# This page documents the path. The static page does not execute FAST-LIVO2.

Expected Result Shape

{
  "tool": "fast-livo2",
  "status": "ok",
  "trajectory": [
    {
      "label": "LiDAR-inertial-visual odometry",
      "score": 0.87,
      "output": "Pose trajectory, local map, reconstruction"
    }
  ],
  "timing": {
    "runtime": "Average split is 17.13 ms LiDAR + 12.90 ms image; ARM average is 78.44 ms. Airborne mapping reports APE RMSE 0.64 m / 0.27 m on two public sequences versus R3LIVE 2.76 m / 0.52 m.",
    "device": "documented in source benchmark when available"
  },
  "artifacts": {
    "visualization": "tools/fast-livo2/runs/visualization.png",
    "raw_predictions": "tools/fast-livo2/runs/predictions.json"
  }
}

Paper figure

Academic Info

Paper identity and contribution summary.

TitleFAST-LIVO2: Fast, Direct LiDAR-Inertial-Visual Odometry

AuthorsChunran Zheng and HKU-MARS collaborators

VenueIEEE Transactions on Robotics, 2024 / arXiv:2408.14035

ContributionFuses LiDAR, inertial, and visual measurements in a direct odometry pipeline to support accurate real-time localization and mapping on robotic platforms.

Citation

@misc{fastlivo22024,
  title={FAST-LIVO2: Fast, Direct LiDAR-Inertial-Visual Odometry},
  author={Chunran Zheng and HKU-MARS collaborators},
  year={2024},
  note={IEEE Transactions on Robotics, 2024 / arXiv:2408.14035},
  url={https://arxiv.org/abs/2408.14035}
}

Benchmark

Only compact, source-reported numbers are shown here.

Dataset	Metric	Value	Runtime	Source
MARS-LVIG AMvalley03	APE RMSE	0.68 m sequential update vs 3.12 m asynchronous and 2.45 m synchronous-standard	30.03 ms average on Intel i7-10700K	FAST-LIVO2 paper
Airborne mapping public sequences	APE RMSE	0.64 m / 0.27 m vs R3LIVE 2.76 m / 0.52 m	17.13 ms LiDAR + 12.90 ms image average	FAST-LIVO2 paper

Artifacts

FAST-LIVO2 paper and supplement, ROS launch files, YAML configs, runtime table, APE/RMSE reports, evaluation logs, and pose trajectories.

Demo Images

Visual references from the original tool. Click any image to inspect the original size.