Reinforcement and Imitation Learning for Diverse Visuomotor Skills
Reinforcement and Imitation Learning for Diverse Visuomotor Skills
复制标题
DOI：
10.15607/rss.2018.xiv.009
复制
发表时间：
2018-02
期刊：
影响因子：
--
通讯作者：
Yuke Zhu;Ziyun Wang;J. Merel;Andrei A. Rusu;Tom Erez;Serkan Cabi;S. Tunyasuvunakool;János Kramár;R. Hadsell;Nando de Freitas;N. Heess
中科院分区：
文献类型：
--
作者：
Yuke Zhu;Ziyun Wang;J. Merel;Andrei A. Rusu;Tom Erez;Serkan Cabi;S. Tunyasuvunakool;János Kramár;R. Hadsell;Nando de Freitas;N. Heess
我们提出了一种无模型的深度强化学习方法，该方法利用少量的演示数据来辅助强化学习代理。我们将这种方法应用于机器人操作任务，并训练端到端的视觉运动策略，这些策略直接从RGB摄像机输入映射到关节速度。我们证明了我们的方法可以解决各种各样的视觉运动任务，对于这些任务，设计一个脚本化控制器将是很费力的。在实验中，我们的强化和模仿代理取得了比单独使用强化学习或模仿学习训练的代理更好的性能。我们还说明，这些策略经过大的视觉和动力学变化训练后，可以在零射击半真实传输中取得初步成功。可在此HTTPS URL中查看该作品的简要视觉描述
We propose a model-free deep reinforcement learning method that leverages a small amount of demonstration data to assist a reinforcement learning agent. We apply this approach to robotic manipulation tasks and train end-to-end visuomotor policies that map directly from RGB camera inputs to joint velocities. We demonstrate that our approach can solve a wide variety of visuomotor tasks, for which engineering a scripted controller would be laborious. In experiments, our reinforcement and imitation agent achieves significantly better performances than agents trained with reinforcement learning or imitation learning alone. We also illustrate that these policies, trained with large visual and dynamics variations, can achieve preliminary successes in zero-shot sim2real transfer. A brief visual description of this work can be viewed in this https URL