Shangshu Yu

Xiamen University, Nanyang Technological University

Papers

Total Citations

412

H-Index

About

Shangshu Yu is a researcher specializing in 3D computer vision, LiDAR-based localization, and deep learning for autonomous systems. His work sits at the intersection of point cloud processing and real-world robotic applications, with a particular focus on enabling precise, efficient localization for autonomous vehicles. Yu's most influential contribution is his comprehensive review of deep learning on 3D point clouds (2020), which has accumulated nearly 370 citations and has become an essential reference for researchers entering the field. This survey systematically examines how deep learning techniques can be applied to point cloud data, reflecting the growing importance of LiDAR sensors in 3D scene understanding. Building on this foundation, Yu has developed several innovative localization frameworks, including STCLoc, which incorporates spatio-temporal constraints into absolute pose regression, and LiDAR-based localization using universal encoding with memory-aware regression. His most recent work, LightLoc (2025), tackles a critical practical bottleneck — dramatically reducing training time for outdoor LiDAR localization systems, making deployment far more feasible for time-sensitive autonomous driving applications. Together, Yu's research demonstrates a consistent commitment to bridging theoretical advances in 3D representation with scalable, real-world localization solutions, establishing him as a rising contributor to autonomous systems research.