Radar4Motion: IMU-Free 4D Radar Odometry with Robust Dynamic Filtering and RCS-Weighted Matching

Codes for paper "Radar4Motion: IMU-Free 4D Radar Odometry with Robust Dynamic Filtering and RCS-Weighted Matching"

Radar4Motion is a robust odometry method that utilizes Doppler and RCS information from the 4D imaging radar's point cloud, even in the presence of noisy and sparse point cloud data.

The code will be uploaded after the review process!

System architecture

Demo

View-of-Delft Dataset Seq 03

View-of-Delft Dataset Seq 17

• The above gif shows ONLY odometry-based mapping results.
  • NO inertial sensor, NO GNSS sensor, NO loop-closure
  • Only Single front-view 4D Imaging Radar!

Contact

If you have any questions, please let me know:

• Soyeong Kim (soyeongkim@hanyang.ac.kr)

Acknowledgement

• In the development of this package, we refer to KISS-ICP and REVE for source codes.
• Dataset: View-of-Delft (VoD)
• Evaluation: evo package for odometry evaluation