Yuewei Fu

Papers

Total Citations

H-Index

About

Yuewei Fu is a roboticist advancing the frontier of 3D semantic mapping by bridging the gap between classical probabilistic methods and modern deep learning. His primary research focuses on developing real-time, uncertainty-aware perception systems for autonomous robots operating in complex environments. Fu’s major contribution is the introduction of Convolutional Bayesian Kernel Inference (ConvBKI), a modular neural network that explicitly updates per-voxel probabilistic distributions within a neural network layer. This approach, detailed in his most-cited paper (14 citations, 2023), combines the mathematical rigor and interpretability of classical Bayesian filters with the efficiency of convolutional architectures. His follow-up work (7 citations, 2024) achieved real-time performance exceeding 10 Hz, a critical milestone for practical deployment. By providing quantifiable uncertainty estimates alongside semantic labels, Fu’s work enables safer and more trustworthy robotic navigation, mapping, and scene understanding. His research directly addresses the cross-roads where modern latent-space methods meet classical, mathematically founded approaches, offering a principled path forward for robust robotic perception in uncertain, real-world settings.