We present DemoStart, a novel auto-curriculum reinforcement learning method capable of learning complex manipulation behaviors on an arm equipped with a three-fingered robotic hand, from only a sparse reward and a handful of demonstrations in simulation. Learning from simulation drastically reduces the development cycle of behavior generation, and domain randomization techniques are leveraged to achieve successful zero-shot sim-to-real transfer. Transferred policies are learned directly from raw pixels from multiple cameras and robot proprioception. Our approach outperforms policies learned from demonstrations on the real robot and requires 100 times fewer demonstrations, collected in simulation. More details and videos in https://sites.google.com/view/demostart.
@article{arxiv.2409.06613,
title = {DemoStart: Demonstration-led auto-curriculum applied to sim-to-real with multi-fingered robots},
author = {Maria Bauza and Jose Enrique Chen and Valentin Dalibard and Nimrod Gileadi and Roland Hafner and Murilo F. Martins and Joss Moore and Rugile Pevceviciute and Antoine Laurens and Dushyant Rao and Martina Zambelli and Martin Riedmiller and Jon Scholz and Konstantinos Bousmalis and Francesco Nori and Nicolas Heess},
journal= {arXiv preprint arXiv:2409.06613},
year = {2024}
}
Comments
15 pages total with 7 pages of appendix. 9 Figures, 4 in the main text and 5 in the appendix