VoxEKF-RIO: A 4D Radar Inertial Odometry Based on Incremental Voxel Map and Iterated Kalman Filter
Jiawei Shen, Chenyu Shen, Zishun Deng, Wanbiao Lin, Bohan Shi, Lei Sun
Abstract
4D mmWave radar provides the point cloud with range, azimuth, elevation, Doppler velocity and operates normally in severe weather conditions. However, due to wavelength characteristics, the noisy and sparse point cloud that 4D radar collects poses great challenges for SLAM research. In this paper, we propose VoxEKF-RIO, a 4D radar inertial odometry. VoxEKF-RIO filters out noisy points and estimates ego-velocity through a preprocessing module and maintains an incremental voxel map to represent the probabilistic models of environments. To improve the accuracy, a reliable scan-to-submap matching method is designed based on the voxel map, using a point filter to obtain valid points with reliable matches, and adopting a distribution-to-distribution matching distance. Iterative Kalman filter is used to fuse radar velocity, point cloud registration, and IMU data for estimating the platform’s motion. The experiments on publicly available 4D radar datasets demonstrate the reliability and high accuracy of VoxEKF-RIO. The ablation studies reveal the benefit of voxel map in describing the environment characteristics and the reliable matching method.
BibTeX
@inproceedings{iros2025_voxekfrioa4drada,
title = {VoxEKF-RIO: A 4D Radar Inertial Odometry Based on Incremental Voxel Map and Iterated Kalman Filter},
author = {Jiawei Shen and Chenyu Shen and Zishun Deng and Wanbiao Lin and Bohan Shi and Lei Sun},
booktitle = {IROS 2025},
year = {2025}
}