DRAP: Dynamics-Risk Adaptive Plasticity for Continual Learning in Physical AI
DRAP: Dynamics-Risk Adaptive Plasticity for Continual Learning in Physical AI
This work introduces DRAP, a replay-free method for continual learning in physical AI that estimates the historical importance of adaptable model tensors using sample-level action gradients weighted by local trajectory-amplification risk.
The paper proposes DRAP, a replay-free continual learning method for physical AI that uses trajectory-amplification risk to adaptively scale learning rates and retain prior behaviors while acquiring new skills on the LIBERO Spatial, Object, and Goal suites.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
Source：arXiv Robotics — research abstracts · arxiv.org