Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

RaUF: Learning the Spatial Uncertainty Field of Radar

Shengpeng Wang, Kuangyu Wang, Wei Wang

TL;DR

This work proposes RaUF, a spatial uncertainty field learning framework that models radar measurements through their physically grounded anisotropic properties and proposes a Bidirectional Domain Attention mechanism that exploits the mutual complementarity between spatial structure and Doppler consistency, effectively suppressing spurious or multipath-induced reflections.

Abstract

Millimeter-wave radar offers unique advantages in adverse weather but suffers from low spatial fidelity, severe azimuth ambiguity, and clutter-induced spurious returns. Existing methods mainly focus on improving spatial perception effectiveness via coarse-to-fine cross-modal supervision, yet often overlook the ambiguous feature-to-label mapping, which may lead to ill-posed geometric inference and pose fundamental challenges to downstream perception tasks. In this work, we propose RaUF, a spatial uncertainty field learning framework that models radar measurements through their physically grounded anisotropic properties. To resolve conflicting feature-to-label mapping, we design an anisotropic probabilistic model that learns fine-grained uncertainty. To further enhance reliability, we propose a Bidirectional Domain Attention mechanism that exploits the mutual complementarity between spatial structure and Doppler consistency, effectively suppressing spurious or multipath-induced reflections. Extensive experiments on public benchmarks and real-world datasets demonstrate that RaUF delivers highly reliable spatial detections with well-calibrated uncertainty. Moreover, downstream case studies further validate the enhanced reliability and scalability of RaUF under challenging real-world driving scenarios.

RaUF: Learning the Spatial Uncertainty Field of Radar

TL;DR

This work proposes RaUF, a spatial uncertainty field learning framework that models radar measurements through their physically grounded anisotropic properties and proposes a Bidirectional Domain Attention mechanism that exploits the mutual complementarity between spatial structure and Doppler consistency, effectively suppressing spurious or multipath-induced reflections.

Abstract

Millimeter-wave radar offers unique advantages in adverse weather but suffers from low spatial fidelity, severe azimuth ambiguity, and clutter-induced spurious returns. Existing methods mainly focus on improving spatial perception effectiveness via coarse-to-fine cross-modal supervision, yet often overlook the ambiguous feature-to-label mapping, which may lead to ill-posed geometric inference and pose fundamental challenges to downstream perception tasks. In this work, we propose RaUF, a spatial uncertainty field learning framework that models radar measurements through their physically grounded anisotropic properties. To resolve conflicting feature-to-label mapping, we design an anisotropic probabilistic model that learns fine-grained uncertainty. To further enhance reliability, we propose a Bidirectional Domain Attention mechanism that exploits the mutual complementarity between spatial structure and Doppler consistency, effectively suppressing spurious or multipath-induced reflections. Extensive experiments on public benchmarks and real-world datasets demonstrate that RaUF delivers highly reliable spatial detections with well-calibrated uncertainty. Moreover, downstream case studies further validate the enhanced reliability and scalability of RaUF under challenging real-world driving scenarios.
Paper Structure (15 sections, 2 theorems, 7 equations, 6 figures, 3 tables)

This paper contains 15 sections, 2 theorems, 7 equations, 6 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Point Cloud Extraction for Radar
  4. Point Cloud Uncertainty Quantification
  5. Methodology
  6. BPM for Radar Uncertainty Learning
  7. Doppler-Aware Predictive Enhancement
  8. Network & Uncertainty Training
  9. Experiments and Evaluation
  10. Dataset
  11. Baseline & Case Studies
  12. Evaluation Metric
  13. Performance and Analysis
  14. Ablation Study
  15. Conclusion

Key Result

Theorem 1

Doppler velocity of the reflection, i.e. relative radial velocity $v^r_{i,j}$, is determined by the radar's ego-velocity $\boldsymbol{v}^r$ and the scatter's directional vector $(\alpha,\beta)$ when the target is stationary relative to the ground, where $(\alpha,\beta)$ denotes the azimuth and eleva

Figures (6)

  • Figure 1: Traditional methods compromise under conflict, leading to ill-posed geometric inference.
  • Figure 2: Our insights: (a) Multi-frame statistical analysis shows that radar PCLs inherently exhibit anisotropic uncertainty, unlike the isotropic nature of LiDAR. (b) The Doppler measurements of positive detections conform to kinematic constraints. (Experimental data are based on Coloradar kramer2022coloradar and RaDelft iroldan2024)
  • Figure 3: Framework of our RaUF for learning radar anisotropic uncertainty with doppler-aware predictive enhancement.
  • Figure 4: Doppler-Consistency for stationary positive scatters.
  • Figure 5: Bird’s-Eye View visualization of Radar point clouds augmented by various models and LiDAR ground truth.
  • ...and 1 more figures

Theorems & Definitions (4)

  • Theorem 1
  • proof
  • Theorem 2
  • proof