Neural dynamics models trained on random simulator data act as a general robot physics backend, supporting policy learning and real-world fine-tuning with stable long-horizon prediction.
Title resolution pending
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
-
Neural Robot Dynamics
Neural dynamics models trained on random simulator data act as a general robot physics backend, supporting policy learning and real-world fine-tuning with stable long-horizon prediction.