Physical robot learning and simulation
Pieter Abbeel
Coauthor of DayDreamer and UniSim, connecting world models with physical learning and interactive simulation.
- Tracked papers
- 4
- Past year
- 2
- Editor picks
- 0
SuperMimic: Scaling Humanoid Whole-Body Control with One Differentiable World Model
SuperMimic trains humanoid whole-body tracking using one learned differentiable world model trained from logged simulator transitions. The world model and goal-conditioned policy train together in a loop; backpropagation-through-time throug…
Counterfactual Video Generation Enables Scalable Humanoid Loco-Manipulation
PRISM uses grounded video-to-video (V2V) generation to create hundreds of counterfactual human–object interaction videos from four real seed videos. A contact-anchored real-to-sim pipeline reconstructs and retargets monocular videos into ph…
Learning Interactive Real-World Simulators
UniSim learns a universal simulator of real-world interaction by orchestrating diverse datasets (image, robotics, navigation) so it can simulate the visual outcome of both high-level instructions and low-level controls; policies trained pur…
DayDreamer: World Models for Physical Robot Learning
DayDreamer applies the Dreamer algorithm to four physical robots that learn online in the real world without simulators, using the same hyperparameters across tasks: a quadruped, two robot arms and a wheeled robot.