NRI: INT: Customizing Semi-Autonomous Nursing Robots Using Human Expertise

遥控机器人有可能让人类在不把自己置于危险之中或不长途跋涉的情况下执行有用的任务。该项目探索人类如何控制护理机器人，这些机器人可以与患者交流，收集生命体征，并在隔离环境中执行日常清洁任务。使用这些机器人有可能在疾病爆发期间保护护士免受感染，并保护免疫系统较弱的患者。这项工作的一个重大挑战是使机器人的用户界面足够容易，护士无需经过大量培训就可以使用。因为工程师不是护理专家，这项研究将让护士通过教机器人关于护理中通常使用的物体、地点和任务来定制用户界面。经过训练后，人工智能算法将自动估计护士想要执行的操作，这些操作将显示在一个简单的用户界面中，允许护士快速选择这些操作。该项目将继续杜克大学工程学院和护理学院之间的跨学科合作。研究将在三个重点领域进行：1)用于监督自治的智能人机界面，学习多模态传感器输入流之间的映射，以提供简单，可解释的任务选项和状态反馈；2)分层任务学习算法，帮助人类专家训练新的复合任务；3)通过注册护士和护生在培训环境中执行模拟临床任务，对人机系统的速度、可靠性、操作员工作量和操作员学习曲线进行真实世界评估。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Remote-controlled robots have the potential to allow humans to perform useful tasks without putting themselves in danger, or without travelling long distances. This project explores how humans can control nursing robots that can communicate with patients, collect vital signs, and perform routine cleaning tasks in quarantine environments. The use of these robots has the potential to protect nurses from infection during disease outbreaks, and to protect patients with weakened immune system. A significant challenge in this effort is to make the user interface to the robot easy enough for nurses to use without significant training. Because engineers are not experts in nursing, the research will let nurses customize the user interface by teaching the robot about objects, places, and tasks that are typically used in nursing. After training, artificial intelligence algorithms will then automatically estimate which actions the nurse wants to perform, and these will be presented in a simple user interface that allows the nurse to select those actions quickly. This project will continue an interdisciplinary collaboration between Duke's School of Engineering and School of Nursing. Research will be conducted in three thrust areas: 1) smart human operator interfaces for supervised autonomy that learn mappings between multimodal sensor input streams to provide simple, interpretable task options and status feedback; 2) hierarchical task learning algorithms for helping human experts train novel composite tasks; and 3) real world evaluation of human-robot system speed, reliability, operator workload, and operator learning curve using registered nurses and nursing students performing simulated clinical tasks in training environments.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.