PDRVLA integrates physical dynamics into robot vision-language-action models
Original titlePDRVLA: A Vision-Language-Action Model with Physical Dynamics Representations
- Outlet
- Hugging Face
- Published
- Importance
- 4/5
- Trending
- 25
Summary
PDRVLA, a vision-language-action model with a lightweight physical dynamics module, generates robot actions from visual input and language instructions. It learns from public data augmented with human-to-robot synthesis and embodied chain-of-thought annotation, achieving 99.3% success on LIBERO and up to 93.6% on RoboTwin 2.0.
Also reported by 1 outlet
Original excerpt
Vision--Language--Action (VLA) models generate robot actions from visual observations and language instructions, but their understanding of physical dynamics remains limited. Video world models can capture scene dynamics, yet generating future videos introduces high latency into the control loop. We introduce PDRVLA, whose lightweight Physical Dynamics Module extracts physical tokens from vision--language features to condition action generation without online video generation. The module learns through direct prediction of future RGB, depth, and optical-flow features, with auxiliary supervision through a frozen pretrained video world model. Training relies entirely on publicly available data, expanded through human-to-robot synthesis, subtask annotation, and embodied chain-of-thought (ECoT) annotation. Across four simulation benchmarks, PDRVLA achieves success rates of 99.3% on LIBERO, 89.9% on LIBERO-Plus, 93.6% on RoboTwin~2.0, and 81.5% on SimplerEnv. Ablations show improved future prediction with world-model supervision and evaluate how subtask conditioning, future object-pose supervision, and ECoT training affect action success.