Papers
6
Total Citations
117
H-Index
4
About
Xinrui Wu is a researcher specializing in 3D computer vision, autonomous systems, and robot localization, with a focus on advancing perception capabilities for autonomous driving and mobile robotics. Wu's most influential contribution, "Hierarchical Attention Learning of Scene Flow in 3D Point Clouds" (2021), has garnered 73 citations and introduced a novel attention-based approach to estimating 3D motion fields from point cloud data — a critical capability for dynamic environment understanding. Building on this foundation, Wu further refined scene flow estimation through context-aware feature extraction in a residual learning framework, accumulating an additional 23 citations across related publications. Beyond scene flow, Wu has made notable contributions to robot localization and odometry. The 2023 work on "Pseudo-LiDAR for Visual Odometry" explores enriching camera-based navigation with LiDAR-like depth representations, while research on GNSS and Visual-Inertial-Wheel Odometry fusion addresses robust, drift-free state estimation for mobile robots. Most recently, Wu's cross-modal localization work using LiDAR heat maps tackles real-world challenges in large-scale scene navigation. Collectively, Wu's research portfolio reflects a consistent drive to bridge sensing modalities and improve spatial understanding in complex, real-world robotic environments.
Research Focus
Key Achievements
Top Papers
- 1Hierarchical Attention Learning of Scene Flow in 3D Point Clouds73 citations · 2021
- 2Residual 3-D Scene Flow Learning With Context-Aware Feature Extraction21 citations · 2022
- 3Pseudo-LiDAR for Visual Odometry14 citations · 2023
- 4
- 5
- 6Residual 3D Scene Flow Learning with Context-Aware Feature Extraction2 citations · 2021