A single locomotion policy trained on 50 simulated robots transfers without fine-tuning to real quadrupeds and to one real humanoid, though a second humanoid needed in-distribution training.
Orbit: A unified simulation framework for interactive robot learning environments,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.RO 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Multi-Embodiment Locomotion at Scale with extreme Embodiment Randomization
A single locomotion policy trained on 50 simulated robots transfers without fine-tuning to real quadrupeds and to one real humanoid, though a second humanoid needed in-distribution training.