Common methods for learning robot dynamics assume motion is continuous, causing unrealistic model predictions for systems undergoing discontinuous impact and stiction behavior. In this work, we resolve this conflict with a smooth, implicit encoding of the structure inherent to contact-induced discontinuities. Our method, ContactNets, learns parameterizations of inter-body signed distance and contact-frame Jacobians, a representation that is compatible with many simulation, control, and planning environments for robotics. We furthermore circumvent the need to differentiate through stiff or non-smooth dynamics with a novel loss function inspired by the principles of complementarity and maximum dissipation. Our method can predict realistic impact, non-penetration, and stiction when trained on 60 seconds of real-world data.
PDF@inproceedings{Pfrommer2020,
arxiv = {2009.11193},
author = {Pfrommer, Samuel and Halm, Mathew and Posa, Michael},
booktitle = {The Conference on Robot Learning (CoRL)},
title = {{ContactNets: Learning Discontinuous Contact Dynamics with Smooth, Implicit Representations}},
year = {2020},
code = {https://github.com/DAIRLab/contact-nets},
youtube = {I6p8JrIp1Es},
url = {https://proceedings.mlr.press/v155/pfrommer21a.html}
}