Xiaodi Yuan
Papers
3
Total Citations
18
H-Index
2
About
Xiaodi Yuan is a leading researcher in embodied AI, robotics simulation, and physics-based computing, whose work is redefining how generalizable robot learning systems are built and evaluated. Yuan’s most significant contribution is the development of **ManiSkill3**, a GPU-parallelized robotics simulation and rendering framework that enables unprecedented compute-scalable approaches to robot learning. By overcoming the limitations of existing simulators—which typically support only narrow task ranges and lack features critical for sim2real transfer—Yuan’s open-source platform allows researchers to train and test embodied agents across diverse, complex scenes at massive scale. The foundational paper on ManiSkill3 (2024) and its demonstration paper (2025) have together garnered over 16 citations in just their first year, signaling rapid adoption by the community. Yuan has also advanced **continuous collision detection (CCD)** with the novel C⁵D algorithm, which uses cone casting to enforce intersection-free constraints in rigid-body simulations far more efficiently than previous primitive-level CCD methods. This work directly addresses the primary performance bottleneck in physics simulation. Through these contributions, Yuan is accelerating the path toward truly generalizable embodied intelligence.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3