End-to-end differentiable physics for learning and control
Author(s)
Smith, Kevin A; Allen, Kelsey Rebecca; Tenenbaum, Joshua B
DownloadPublished version (794.1Kb)
Terms of use
Metadata
Show full item recordAbstract
© 2018 Curran Associates Inc.All rights reserved. We present a differentiable physics engine that can be integrated as a module in deep neural networks for end-to-end learning. As a result, structured physics knowledge can be embedded into larger systems, allowing them, for example, to match observations by performing precise simulations, while achieves high sample efficiency. Specifically, in this paper we demonstrate how to perform backpropagation analytically through a physical simulator defined via a linear complementarity problem. Unlike traditional finite difference methods, such gradients can be computed analytically, which allows for greater flexibility of the engine. Through experiments in diverse domains, we highlight the system's ability to learn physical parameters from data, efficiently match and simulate observed visual behavior, and readily enable control via gradient-based planning methods. Code for the engine and experiments is included with the paper.
Date issued
2018-12Department
Massachusetts Institute of Technology. Department of Brain and Cognitive SciencesJournal
32nd Conference on Neural Information Processing Systems (NeurIPS 2018)
Publisher
Curran Associates Inc
Citation
Belbute-Peres, Filipe de A. et al. “End-to-end differentiable physics for learning and control.” Paper presented at the 32nd Conference on Neural Information Processing Systems (NeurIPS 2018), Montréal, Dec 3-8 2018, Curran Associates Inc © 2018 The Author(s)
Version: Final published version