Presentation

<
SMP: Reusable Score-Matching Motion Priors for Physics-Based Character Control
SessionMotion Generation, Warping & Control
DescriptionSMP builds reusable and modular reward models for training motor controllers. Once constructed from a motion dataset, the priors can be reused across diverse tasks while preserving the behaviors in the data, without requiring access to the original dataset or retraining.
Authors
Yuxuan Mu
Simon Fraser University
Ziyu Zhang
Simon Fraser University
Yi Shi
Simon Fraser University
Dun Yang
Simon Fraser University
Minami Matsumoto
Sony Interactive Entertainment
Kotaro Imamura
Sony Interactive Entertainment
Guy Tevet
Stanford University
Chuan Guo
Snap
Michael Taylor
Sony Interactive Entertainment
Chang Shu
National Research Council Canada
Pengcheng Xi
National Research Council Canada
Xue Bin Peng
Simon Fraser University
NVIDIA
Event Type
Technical Paper
TimeTuesday, 21 July 20262:10pm - 2:20pm PDT
LocationRoom 403 A
Digital Library PDF
Session Time & Location
Sunday, 19 July 20266:00pm - 8:45pm PDTHall K
Tuesday, 21 July 20262:00pm - 3:30pm PDTRoom 403 A
Keywords
Animation
Artificial Intelligence/Machine Learning
Simulation
Registration Categories
Full Conference Supporter
Full Conference
Next PresentationNext PresentationMotionBricks: Scalable Real-Time Motions with Modular Latent Generative Model and Smart Primitives
Similar Presentations
GPC: Large-Scale Generative Pretraining for Transferable Motor Control
MOCHI: Motion Enhancement of Collaborative Human-object Interactions
MotionBricks: Scalable Real-Time Motions with Modular Latent Generative Model and Smart Primitives