自定义正弦波足以用于不同风格双足步态的模仿学习
机器人学
2022-05-17 v2
摘要
直到最近，鲁棒的双足 locomotion 才通过强化学习实现。然而，现有实现严重依赖人类专家的见解和努力，这对于机器人系统的迭代设计而言成本高昂。并且，所学运动的风格严格受限于参考对象的风格。在本文中，我们提出一种新方法，以简单的正弦波作为足部高度参考来学习双足 locomotion。基于两足应交替且周期性抬起的朴素人类见解，我们在Cassie机器人上实验证明，一个简单的奖励函数能够使机器人端到端且高效地学会行走，而无需任何显式的模型知识。通过自定义正弦波，所学步态模式也可具有定制化风格。代码发布于 github.com/WooQi57/sin-cassie-rl。
引用
@article{arxiv.2204.04157,
title = {Custom Sine Waves Are Enough for Imitation Learning of Bipedal Gaits with Different Styles},
author = {Qi Wu and Chong Zhang and Yanchen Liu},
journal= {arXiv preprint arXiv:2204.04157},
year = {2022}
}
备注
\c{opyright} 2022 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works