RaDiVe: Robust 4D Radar Odometry with Distance-Bounded NDT and Velocity-Discrepancy Point Uncertainty

Researchers introduced RaDiVe, a 4D radar odometry framework for robot localization.

It combines distance-bounded NDT, a Doppler velocity-discrepancy uncertainty model, and implicit neural mapping to build cleaner local submaps.

On public datasets, RaDiVe achieved lower translation and rotation errors than prior 4D radar odometry methods while keeping real-time performance.

The approach could help robots navigate more reliably in rain, fog, and other conditions where cameras or lidar struggle.