search

zijiechenrobotics / ig_lio

iG-LIO: An Incremental GICP-based Tightly-coupled LiDAR-inertial Odometry
GNU General Public License v2.0
355 stars 48 forks source link

About Performance in High Speed Sports #10

Closed 19855179496 closed 5 months ago

19855179496 commented 5 months ago

May I ask how this algorithm performs in high-speed motion (20m/s), what is its pose error, and is its accuracy sufficient...

zijiechenrobotics commented 5 months ago

Thank you for your interest in this work. This is a good question! I verified iG-LIO's accuracy on the Apollo dataset: https://developer.apollo.auto/daoxianglake.html. However, this experiment was removed due to the page limitation of RA-L.

The speed of the dataset was around 10~14 m/s, not 20m/s. The results are listed as follows:

RPE per100 meters and time(ms) Seq iG-LIO(RPE / time) Faster-LIO FastLIO2 Dist(km)
daoxianglake_20190924124848 0.096 / 18.12 0.109 / 24.71 0.101 / 46.42 47.973
daoxianglake_20191014142530 0.097 / 18.88 0.102 / 25.33 0.097 / 47.31 27.825
daoxianglake_20191021162130 0.092 / 18.45 0.096 / 26.24 0.087 / 48.00 41.013
daoxianglake_20191025104732 0.092 / 18.42 0.094 / 26.06 0.092 / 47.41 18.734
daoxianglake_20191225153609 0.076 / 18.52 0.086 / 28.04 0.087 / 47.40 21.45

I think iG-LIO can work well as long as the following conditions are met:

  1. The movement of the sensing system should not exceed the range of the IMU
  2. The environment is not extremely narrow or lack of geometric features.

I hope these experiment results and suggestions will help you :smile: