This research paper explores uncertainty quantification for automotive radar target detection using two deep learning frameworks: a von Mises (VM) ensemble and an evidential deep learning (EDL) approach. The VM ensemble provides interpretable uncertainty measures aligned with directional geometry, while EDL offers smoother uncertainty variations. Both methods were evaluated on their ability to estimate the direction of arrival (DOA) for radar targets under various conditions. The study found that the VM ensemble is more sensitive to severe perturbations and allows for direct probabilistic integration into tracking systems, highlighting a trade-off between geometric consistency and statistical generality in uncertainty estimation. AI
RANK_REASON The item is a research paper detailing a novel approach to uncertainty quantification in automotive radar. [lever_c_demoted from research: ic=1 ai=0.7]
Read on Hugging Face Daily Papers →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
