arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2007.06045cs.ROcs.LGcs.SYeess.SY
收藏

用神经网络增强可微模拟器以缩小Sim2Real差距

Augmenting Differentiable Simulators with Neural Networks to Close the Sim2Real Gap

University of Southern California(南加州大学);Google(谷歌)

Eric Heiden, David Millard, Erwin Coumans, Gaurav S. Sukhatme

摘要:

我们提出了一个可微的仿真架构铰接刚体动力学,使增强的分析模型与神经网络在任何一点的计算。通过基于梯度的优化,识别的仿真参数和网络权重进行有效的初步实验,在现实世界的数据集和sim2sim传输应用程序,而穷人的局部最优值通过随机搜索方法克服。

英文摘要:

We present a differentiable simulation architecture for articulated rigid-body dynamics that enables the augmentation of analytical models with neural networks at any point of the computation. Through gradient-based optimization, identification of the simulation parameters and network weights is performed efficiently in preliminary experiments on a real-world dataset and in sim2sim transfer applications, while poor local optima are overcome through a random search approach.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑