Elastic-WAM: Adaptive Spatiotemporal Imagination for Efficient Robot Control
Elastic-WAM: Adaptive Spatiotemporal Imagination for Efficient Robot Control
This research paper proposes Elastic-WAM, a novel approach for efficient robot control that adaptively allocates world refinement computation across temporal denoising and spatial future space. By parameterizing a unified spatiotemporal refinement demand field learned through action-grounded supervision, the method guides recurrent sparse world refinement to improve task success and reduce inference latency. Evaluations on diverse simulation benchmarks and real-world long-horizon manipulation tasks demonstrate state-of-the-art control performance with reduced computational overhead compared to prior iterative video denoising approaches.
The paper introduces Elastic-WAM, a method that adaptively allocates computation for world refinement in video-based robot control to reduce inference overhead while maintaining state-of-the-art performance on simulation and real-world manipulation benchmarks.
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