Enabling Robot Teammates to Learn Latent States of Human Collaborators
Enabling Robot Teammates to Learn Latent States of Human Collaborators
复制标题
使机器人队友能够了解人类协作者的潜在状态
DOI：
--
复制
发表时间：
2010
期刊：
影响因子：
--
通讯作者：
J. Shah
中科院分区：
文献类型：
--
作者：
Vaibhav Unhelkar;Charlie Guan;N. Roy;J. Shah
我们有兴趣设计协作机器人，可以无缝地在复杂的领域，如协同制造和灾难响应与人类队友进行交互。为了成为成功的队友，这些机器人需要能够建模，预测和适应人类合作者。然而，对人类队友的行为进行建模是具有挑战性的，因为人类的决策通常取决于潜在的且难以指定的因素[4]。在这个扩展的摘要中，我们描述了为设计机器人协作者而对人类建模的挑战，总结了我们对这个问题的算法解决方案，最后描述了一个旨在评估我们方法的人机协作场景。
We are interested in designing collaborative robots that can seamlessly interact in complex domains such as, collaborative manufacturing and disaster response with human teammates. To be successful teammates, such robots need the ability to model, predict and adapt to their human collaborators. However, modeling the behavior of human teammates is challenging since human decisions often depend on factors that are latent and difficult to specify [4]. In this extended abstract, we describe this challenge for modeling humans for designing robot collaborators, summarize our algorithmic solutions towards this problem and conclude with a description of a human-robot collaboration scenario designed to evaluate our approach.