RaUF: Learning the Spatial Uncertainty Field of Radar
This work addresses the challenge of ambiguous feature-to-label mapping in radar perception for autonomous driving, offering enhanced reliability in adverse conditions, though it appears incremental as it builds on existing cross-modal supervision methods.
The paper tackles the problem of low spatial fidelity and ambiguity in millimeter-wave radar for perception tasks by proposing RaUF, a spatial uncertainty field learning framework that models radar measurements with anisotropic properties, resulting in highly reliable spatial detections with well-calibrated uncertainty as validated on public benchmarks and real-world datasets.
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.