$\Sigma$-LeWM: Learning World Models via S-Information for Robot Control
$\Sigma$-LeWM: Learning World Models via S-Information for Robot Control
This paper introduces $\Sigma$-LeWM, an extension of the LeWorldModel architecture that incorporates backward dynamics and inverse dynamics action predictors alongside a latent forward dynamics predictor. The authors argue that these additional losses enforce further structure on the latent embedding, improving performance and planning efficiency for real-time model-based control. Results are demonstrated on standard world model benchmarks and two simulated robot manipulation tasks.
The authors propose adding backward dynamics and inverse dynamics action predictors to the LeWorldModel architecture to improve latent structure for real-time control, with evaluations reported on standard benchmarks and simulated manipulation tasks.
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