Eliot Xing Vernon Luk Jean Oh

Carnegie Mellon University

TL;DR We present a maximum entropy first-order model-based RL algorithm, alongside a parallel differentiable multiphysics simulation platform for RL that supports simulating various materials beyond just rigid bodies.

Results on Rewarped tasks

BibTeX

@article{xing2024stabilizing,
    title={Stabilizing Reinforcement Learning in Differentiable Multiphysics Simulation},
    author={Eliot Xing and Vernon Luk and Jean Oh},
    journal={arXiv preprint arXiv:2412.12089},
    year={2024}
}