New Denoiser Improves UWB Sensing for Work-Zone Reconstruction.

Ultra-wideband (UWB) sensing offers a cost-effective way to perceive work-zone geometry, which is crucial for intelligent transportation systems. However, its accuracy is often hampered by issues like non-line-of-sight propagation, burst noise, and long-tail errors, which can distort spatial reconstruction. Researchers have developed GAIA, a novel geometry-aware, infrastructure-anchored learning framework designed to address these challenges. GAIA integrates temporal range modeling with the estimation of latent anchor layouts and deterministic distance projection. Its core innovation lies in preserving range denoising as a primary task while ensuring the learned distances contribute to a consistent boundary reconstruction. Evaluations on real-world outdoor UWB datasets, including stress tests, show GAIA's superior performance. It achieved the lowest range mean squared error and the highest polygon Intersection over Union compared to other filtering and learning-based baselines, demonstrating a significant improvement in spatial coherence for work-zone reconstruction.