HCC: Small: Modeling Ad Hoc Collaboration on Complex Manipulation Tasks for Human-Robot Teams

该项目将研究人们如何相互协作来完成复杂的物理任务，并研究使机器人能够与人类协作完成类似任务的方法。许多任务协作都是临时出现的，团队成员在任​​务展开时决定谁负责任务的哪些部分。在人机团队合作中实现类似的紧急协作对于家庭援助、灵活制造和搜索救援等应用至关重要。该项目将有助于理解复杂的人类团队合作以及建立有效的人机团队所需的基本技术。这项研究将得到广泛传播，并将通过指导具有不同背景的学生团队来解决项目的人员和技术方面所需的问题，从而扩大对计算的参与。为了实现这些目标，该项目将调查复杂操作任务的临时协作。虽然多年来对人机协作的研究不断扩大，但之前的工作很大程度上局限于小规模任务（例如桌面操作）、二元团队合作（即两人团队）和/或规定协作（例如具有预定角色和程序的任务）的研究。通过这个项目，研究团队将1）研究一个原则框架，用于描述、表示、推理和实现基于复杂操作任务上人类临时协作的严格建模的协作交互； 2）开发使用感知数据推断和基础协作事件和行动的计算方法； 3) 探索如何利用已开发的框架和推理能力在实验受限的环境中实现临时人机协作。该奖项反映了 NSF 的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。

This project will study how people collaborate with each other to complete complex physical tasks and investigate ways to enable robots to collaborate with people in similar tasks. Many task collaborations emerge on an ad hoc basis, with teammates making decisions about who does what parts of the task as it unfolds. Achieving similar emergent collaboration in human-robot teamwork is critical for applications such as in-home assistance, flexible manufacturing, and search-and-rescue. This project will contribute to the understanding of complex human teamwork and the fundamental technologies needed to bring about effective human-robot teams. This research will be widely disseminated, and will broaden participation in computing by mentoring teams of students with the diverse backgrounds needed to address both the human and the technical aspects of the project.To achieve these goals, this project will investigate ad hoc collaboration on complex manipulation tasks. While research on human-robot collaboration has expanded over the years, prior work has been largely limited to the investigation of small-scale tasks (e.g., tabletop manipulation), dyadic teamwork (i.e., two-person teams), and/or prescribed collaboration (e.g., tasks with predetermined roles and procedures). Through this project, the research team will 1) investigate a principled framework for describing, representing, reasoning, and enabling collaborative interactions based on rigorous modeling of human ad hoc collaboration on complex manipulation tasks; 2) develop computational methods for inferring and grounding collaborative events and actions using perception data; and 3) explore how the developed framework and reasoning capabilities may be leveraged to enable ad hoc human-robot collaboration in an experimentally constrained setting.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.