NRI: FND: Communicating Physical Interactions

为了让机器人成为无处不在的合作者，它们必须在非结构化的人类环境中与物体进行物理交互。机器人必须熟练地抓住、推动、挤压、折断、平衡、稳定和移交物体，并且必须与人类合作者有效地沟通这些动作。人们不仅必须能够指定机器人做什么和如何做，他们还必须能够解释机器人打算做什么（以及如何做）。如果机器人要执行诸如为老年人提供护理、响应性地帮助技术人员进行维修或接受非专业人员的培训以执行重复性组装任务等任务，那么关于物理交互的有效沟通将是必不可少的。该项目将实现新一代机器人应用，并通过开发更好的方法使机器人更有效地与人沟通，推进无处不在的协作机器人的愿景。该项目将解决关于物理交互的沟通中的关键挑战：传达不可见和不熟悉的数量（例如，力量和遵从），沟通计划和突发事件，以及沟通没有（或不应该）发生的事情。该项目将涉及三个关键挑战：规范、解释和监控。为了应对这些挑战，该项目将1)进行形成性研究，以深入了解人们如何沟通身体互动和解释展示；2)开发基于增强演示思想的指定物理动作的方法，基于可解释表示思想的解释物理动作的方法，以及基于多模态通信的监测物理动作的方法；3)将这些想法部署到原型系统中，以进行情境化的评估。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。

For robots to become ubiquitous collaborators, they must interact physically with objects in unstructured human environments. Robots must adeptly grasp, push, squeeze, snap, balance, stabilize and hand-off objects, and must effectively communicate about these actions with their human collaborators. People must not only be able to specify to a robot what to do and how to do it, they must also be able to interpret what a robot intends to do (and how). Effective communication about physical interactions will be essential if robots are going to perform such tasks as delivering care to older adults, responsively helping technicians in performing repairs, or being trained by non-experts to perform repetitive assembly tasks. This project will enable a new generation of robot applications, and advance the vision of ubiquitous, collaborative robots by developing better methods for robots to more effectively communicate with people. The project will address key challenges in communication about physical interactions: conveying invisible and unfamiliar quantities (e.g., forces and compliances), communicating plans and contingencies, and communicating about what did not (or should not) happen. The project will involve three key challenges: specification, interpretation, and monitoring. To address these challenges, the project will 1) perform formative studies to gain insight into how people communicate about physical interactions and interpret displays; 2) develop methods for specifying physical actions based on the idea of augmented demonstrations, methods for interpreting physical actions based on the idea of interpretable representations, and methods for monitoring physical actions based on multimodal communication; and 3) deploy these ideas in prototype systems for contextualized scenarios for evaluation.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.