VGC-RIO: A Tightly Integrated Radar-Inertial Odometry with Spatial Weighted Doppler Velocity and Local Geometric Constrained RCS Histograms

TL;DR

VGC-RIO addresses robust radar–inertial odometry for sparse, noisy 4D radar data by tightly coupling IMU pre-integration with three residuals: $r_I$, $w_D r_D$, and $r_P$. It introduces a spatial-distribution based Doppler weighting scheme and a Local Geometric constrained RCS (LGC) histogram descriptor for robust keypoint registration, along with Neighborhood-Expanded Histogram Intersection (NHI) and RANSAC-based outlier rejection. The method is validated on public and self-constructed datasets, showing improved accuracy and robustness over state-of-the-art radar–inertial methods under aggressive motion and sparse point clouds, with competitive performance relative to lidar in some scenarios. The work advances radar–inertial fusion by leveraging Doppler distribution and RCS-driven local geometry for reliable localization in GNSS-denied, adverse conditions, and suggests future loop-closure possibilities via RCS features.

Abstract

Recent advances in 4D radar-inertial odometry have demonstrated promising potential for autonomous lo calization in adverse conditions. However, effective handling of sparse and noisy radar measurements remains a critical challenge. In this paper, we propose a radar-inertial odometry with a spatial weighting method that adapts to unevenly distributed points and a novel point-description histogram for challenging point registration. To make full use of the Doppler velocity from different spatial sections, we propose a weighting calculation model. To enhance the point cloud registration performance under challenging scenarios, we con struct a novel point histogram descriptor that combines local geometric features and radar cross-section (RCS) features. We have also conducted extensive experiments on both public and self-constructed datasets. The results demonstrate the precision and robustness of the proposed VGC-RIO.

Figures (6)

• Figure 1: Upper left: Vehicle platform and sensors. Upper right: Point cloud segmented into a spherical coordinate system-like structure, with Doppler weights and LGC histograms extracted (details in later sections). Dynamic points shown in black. Lower: Trajectory and mapping results visualized on Google Earth.
• Figure 2: System overview of VGC-RIO
• Figure 3: Visualization of weights at different points and in different directions. The point cloud is colored according to the weights.
• Figure 4: The LGC histograms of corresponding points between consecutive frames. For the sake of visualization, the distance dimension and the RCS dimension of the 2D histogram are compared separately.
• Figure 5: The trajectories of four systems on the Snail-Radar dataset