Collective Learning for Unified Robotic Manipulation via Predictive Dynamics Adaptation
Collective Learning for Unified Robotic Manipulation via Predictive Dynamics Adaptation
This research proposes PACE, a two-stage collective learning framework that separates shared task-space intent from embodiment-dependent transition dynamics. The first stage uses a pre-trained representation layer to extract task-level context, while the second stage introduces a predictive dynamics adapter that grounds policies in the executing robot's transition behavior. Evaluations show PACE achieves over 93% success in simulation and validates sim-to-real transfer with a physical KUKA robot.
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.
- Environment
- SimulationOpen source S4
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
level context from collective data. The second stage introduces a decision layer equipped with a predictive dynamics adapter. By forcing the latent state to predict future transitions, this adapter acts as a latent regularizer that grounds the policy in the transition behavior of the executing robot while preserving real-time inference without online planning. Evaluations across simulated embodiments demonstrate that
Open source S4
our method prevents performance collapse on distinct robots. It achieves over 93\% overall success and 89\% on dynamically out-of-distribution zero-shot simulated robots, outperforming the Transformer behavior cloning (82\%) and diffusion policy baselines (66\%). Physical KUKA deployment further validates the feasibility of the full state-based pipeline under zero-shot sim-to-real transfer.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org