Zhengrong Xue

Shanghai Jiao Tong University, Tsinghua University

Papers

Total Citations

H-Index

About

Zhengrong Xue is a researcher advancing the frontiers of 3D computer vision and robotic manipulation. His work centers on developing unsupervised, alignment-free methods for keypoint detection, a critical challenge for enabling robots to generalize across unseen objects and poses. Xue’s foundational contribution, "Skeleton Merger" (2021, 35 citations), introduced an unsupervised aligned keypoint detector, eliminating the need for labor-intensive, ambiguous keypoint annotations. This breakthrough directly enabled his later work, "USEEK" (2023, 13 citations), which achieves SE(3)-equivariant keypoint detection for generalizable manipulation. USEEK allows a robot to grasp intra-category objects in arbitrary poses from a single demonstration, a significant leap toward practical, data-efficient robotics. Most recently, Xue introduced "ArrayBot" (2024, 12 citations), a novel distributed manipulation system integrating a 16×16 array of tactile-sensor-equipped pillars. This work tackles the grand challenge of generalizable, touch-based distributed manipulation, promising more dexterous and perceptive robotic interactions. Through these contributions, Xue is systematically building the algorithmic and hardware foundations for robots that can perceive, understand, and manipulate the physical world with unprecedented autonomy and adaptability.