Zhenjia Xu
Papers
9
Total Citations
638
H-Index
6
About
Zhenjia Xu is a robotics researcher whose work sits at the intersection of robot learning, manipulation, and physical scene understanding. He is best known for his landmark contribution to **Diffusion Policy** (2024), which reframes visuomotor robot control as a conditional denoising diffusion process — a method benchmarked across 15 tasks and 4 manipulation suites, accumulating over 338 citations and rapidly becoming a foundational technique in robot learning. His **Universal Manipulation Interface** (UMI) further demonstrates his commitment to scalable, practical robotics, offering a framework that transfers skills from in-the-wild human demonstrations directly to deployable robot systems, garnering over 131 citations. Earlier in his career, Xu tackled the challenge of physical object understanding through **DensePhysNet** (2019), showing that robots could learn dense physical representations via dynamic interactions — work that attracted over 87 citations. His portfolio also includes **AdaGrasp**, advancing gripper-agnostic grasping policies, and research on 3D dynamic scene representations and cross-embodiment skill discovery. Across these contributions, Xu has established himself as an influential voice shaping the future of generalizable, learning-driven robot manipulation.
Research Focus
Key Achievements
Top Papers
- 1Diffusion policy: Visuomotor policy learning via action diffusion338 citations · 2024
- 2
- 3
- 4AdaGrasp: Learning an Adaptive Gripper-Aware Grasping Policy34 citations · 2021
- 5Learning 3D Dynamic Scene Representations for Robot Manipulation21 citations · 2020
- 6
- 7
- 8AdaGrasp: Learning an Adaptive Gripper-Aware Grasping Policy3 citations · 2020
- 9XSkill: Cross Embodiment Skill Discovery3 citations · 2023