Learning Representations that Enable Generalization in Assistive Tasks
Learning Representations that Enable Generalization in Assistive Tasks
复制标题
能够泛化辅助任务的学习表示
DOI：
10.48550/arxiv.2212.03175
复制
发表时间：
2022
期刊：
影响因子：
--
通讯作者：
A. Dragan
中科院分区：
文献类型：
--
作者：
Jerry Zhi;Aditi Raghunathan;Daniel S. Brown;Zackory M. Erickson;A. Dragan
sim2real最近的工作已经成功地使机器人能够在物理环境中通过不同的“人口”环境（即领域随机化）进行模拟训练。在这项工作中，我们专注于在辅助任务中实现泛化：机器人协助用户的任务（例如，帮助有运动障碍的人洗澡或搔痒）。与之前的模拟现实成功相比，这样的任务特别有趣，因为现在的环境中包含了一个也在演戏的人。这使问题复杂化，因为人类用户的多样性（而不仅仅是物理环境参数）更难以在种群中捕获，从而增加了在测试时遇到分布外（OOD）人类策略的可能性。我们主张对这种OOD策略的泛化受益于(1)学习人类策略的良好潜在表示，可以准确地映射到测试时间的人类，以及(2)使该表示与测试时间交互数据相适应，而不是依赖于它来完全捕获基于模拟人口的人类策略空间。我们通过评估有目的构建的OOD测试策略来研究如何最好地学习这种表示。我们发现编码环境（或人口）参数的sim2real方法在机器人孤立完成的任务中工作得很好，但在辅助任务中工作得不好。在辅助方面，根据交互历史直接训练表征似乎至关重要，因为这是机器人在测试时可以访问的内容。此外，训练这些表征来预测人类的行为，不仅可以给它们提供更好的结构，还可以在测试时对它们进行微调，当机器人观察同伴的行为时。https://adaptive-caregiver.github.io。
Recent work in sim2real has successfully enabled robots to act in physical environments by training in simulation with a diverse ''population'' of environments (i.e. domain randomization). In this work, we focus on enabling generalization in assistive tasks: tasks in which the robot is acting to assist a user (e.g. helping someone with motor impairments with bathing or with scratching an itch). Such tasks are particularly interesting relative to prior sim2real successes because the environment now contains a human who is also acting. This complicates the problem because the diversity of human users (instead of merely physical environment parameters) is more difficult to capture in a population, thus increasing the likelihood of encountering out-of-distribution (OOD) human policies at test time. We advocate that generalization to such OOD policies benefits from (1) learning a good latent representation for human policies that test-time humans can accurately be mapped to, and (2) making that representation adaptable with test-time interaction data, instead of relying on it to perfectly capture the space of human policies based on the simulated population only. We study how to best learn such a representation by evaluating on purposefully constructed OOD test policies. We find that sim2real methods that encode environment (or population) parameters and work well in tasks that robots do in isolation, do not work well in assistance. In assistance, it seems crucial to train the representation based on the history of interaction directly, because that is what the robot will have access to at test time. Further, training these representations to then predict human actions not only gives them better structure, but also enables them to be fine-tuned at test-time, when the robot observes the partner act. https://adaptive-caregiver.github.io.
DOI：
10.1613/jair.1.13326
发表时间：
2021-07
期刊：
J. Artif. Intell. Res.
影响因子：
--
作者：
Yuexiang Zhai;Christina Baek;Zhengyuan Zhou;Jiantao Jiao;Yi Ma
通讯作者：
Yuexiang Zhai;Christina Baek;Zhengyuan Zhou;Jiantao Jiao;Yi Ma
DOI：
--
发表时间：
2020-11
期刊：
ArXiv
影响因子：
--
作者：
Annie Xie;Dylan P. Losey;R. Tolsma;Chelsea Finn;Dorsa Sadigh
通讯作者：
Annie Xie;Dylan P. Losey;R. Tolsma;Chelsea Finn;Dorsa Sadigh