Presentation / Installation


Composite Motion Learning With Task Control
SessionCharacter Animation
DescriptionWe present a multi-objective learning approach for composite and task-driven motion control for physically simulated characters. Without needing to manually blend reference motions, our approach learns composite motions directly from multiple reference sources across distinct body parts and supports sample-efficient training from pretrained controllers in an incremental manner.
Authors
Pei Xu
Clemson University
Roblox
Xiumin Shang
University of California Merced
Victor Zordan
Roblox
Clemson University
Ioannis Karamouzas
Clemson University
Event Type
Technical Paper
TimeWednesday, 9 August 202310:45am - 10:56am PDT
LocationPetree Hall C
ACM Digital Library Technical Papers pdfs
Session Time & Location
Sunday, 6 August 20236pm - 8:30pm PDTWest Hall B
Wednesday, 9 August 202310:45am - 12:15pm PDTPetree Hall C
Interest Areas
Research & Education
Recordings
Livestreamed
Recorded
Keywords
Animation/Simulation
Artificial Intelligence/Machine Learning
Registration Categories
FC
FCS
V
VS
EFC
Next PresentationNext PresentationExample-based Motion Synthesis via Generative Motion Matching