Presentation

<
Policy-Space Diffusion for Physics-Based Character Animation
SessionPhysics-Based Human Characters
DescriptionWe present a new perspective on physics-based character animation. Assuming policies for similar motions should have similar weights, we introduce regularization during RL training to preserve weight similarity. By modeling the weights’ manifold with a diffusion model, we generate a continuum of policies adapting to novel character morphologies and tasks.
Authors
Michele Rocca
University of Copenhagen
Sune Darkner
University of Copenhagen
Kenny Erleben
University of Copenhagen
Sheldon Andrews
École de Technologie Supérieure (ÉTS)
Presenter
Michele Rocca
University of Copenhagen
Event Type
Technical Paper
TimeWednesday, 13 August 20253:45pm - 3:55pm PDT
LocationWest Building, Rooms 220-222
Session Time & Location
Sunday, 10 August 20256:00pm - 8:45pm PDTWest Building, Ballroom AB
Wednesday, 13 August 20253:45pm - 5:35pm PDTWest Building, Rooms 220-222
Recordings
Livestreamed
Recorded
Next PresentationNext PresentationMAGNET: Muscle Activation Generation Networks for Diverse Human Movement
Similar Presentations
Diffuse-CLoC: Guided Diffusion for Physics-based Character Look-ahead Control
Multi-Person Interaction Generation from Two-Person Motion Priors
AMOR: Adaptive Character Control through Multi-Objective Reinforcement Learning