LEARNING WITH DIFFERENTIABLE PHYSICS PRIORS

Yiling Qiao

Gensis AI



*** ABSTRACT ***
We develop scalable and efficient differentiable physics simulators for rigid bodies, cloth, fluids, articulated systems, and deformable solids. Our methods achieve orders-of-magnitude speedups over existing approaches while maintaining accuracy and flexibility. These simulators enable solving inverse problems, training control policies, and advancing reinforcement learning.


*** BIOGRAPHY ***
Yiling Qiao earned a Ph.D. in Computer Science from the University of Maryland, College Park, advised by Prof. Ming Lin, and was a member of the GAMMA Group. Her research focuses on physically based simulation, computer graphics, and machine learning, supported by the Meta PhD Fellowship (AR/VR Computer Graphics Track). He received the Larry S. Davis Dissertation Award and is currently working at Genesis AI on physics simulation and general-purpose robotics.