Graph network simulators can learn discontinuous, rigid contact dynamics
Graph network simulators can learn discontinuous, rigid contact dynamics
复制标题
图网络模拟器可以学习不连续的刚性接触动力学
DOI：
--
复制
发表时间：
2022
期刊：
影响因子：
--
通讯作者：
T. Pfaff
中科院分区：
文献类型：
--
作者：
Kelsey R. Allen;Tatiana Lopez;Yulia Rubanova;Kimberly L. Stachenfeld;Alvaro Sanchez;P. Battaglia;T. Pfaff
：近年来，使用深度学习对不连续动力学进行建模的技术不断涌现，例如刚性接触或切换运动模式。一个常见的说法是，由于深度网络参数化方式的连续性，如果没有用于处理接触的显式模块，深度网络就无法准确地建模刚体动力学。在这里，我们通过对已建立的真实和模拟数据集进行实验来研究这一主张，并表明通用图网络模拟器在没有特定于接触的假设的情况下可以学习和预测接触不连续性。此外，图网络模拟器学习到的接触动力学比高度工程化的机器人模拟器更准确地捕获现实世界的立方体投掷轨迹，即使只提供 8 - 16 个轨迹。总的来说，这表明刚体动力学不会对具有适当通用架构和参数化的深度网络构成根本挑战。相反，我们的工作开辟了新的方向，以考虑基于深度学习的模型何时可能比传统的模拟环境更好，以准确地模拟现实世界的接触动力学。
: Recent years have seen a rise in techniques for modeling discontinuous dynamics, such as rigid contact or switching motion modes, using deep learning. A common claim is that deep networks are incapable of accurately modeling rigid-body dynamics without explicit modules for handling contacts, due to the continuous nature of how deep networks are parameterized. Here we investigate this claim with experiments on established real and simulated datasets and show that general-purpose graph network simulators, with no contact-specific assumptions, can learn and predict contact discontinuities. Furthermore, contact dynamics learned by graph network simulators capture real-world cube tossing trajectories more accurately than highly engineered robotics simulators, even when provided with only 8 – 16 trajectories. Overall, this suggests that rigid-body dynamics do not pose a fundamental challenge for deep networks with the appropriate general architecture and parameterization. Instead, our work opens new directions for considering when deep learning-based models might be preferable to traditional simulation environments for accurately modeling real-world contact dynamics.
DOI：
10.1109/iros51168.2021.9636383
发表时间：
2021-03
期刊：
2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)
影响因子：
--
作者：
Mihir Parmar-;Mathew Halm;Michael Posa
通讯作者：
Mihir Parmar-;Mathew Halm;Michael Posa
影响因子：
12.9
作者：
Nicholls, Thomas P.;Constable, Grace E.;Bissember, Alex C.
通讯作者：
Bissember, Alex C.
ContactNets：通过平滑、隐式表示学习不连续接触动力学
DOI：
--
发表时间：
2020
期刊：
Conference on Robot Learning
影响因子：
--
作者：
Pfrommer, Samuel;Halm, Mathew;Posa, Michael
通讯作者：
Posa, Michael