CAREER: Modeling Group Human-Robot Interactions: Towards A Unified Data-Driven Perspective

学院早期职业发展(Career)基金的长期研究目标是促进人-机器人交互(HRI)，使机器人能够有效地参与与多个用户的社交接触。虽然社会机器人研究传统上专注于一对一的互动，但现实世界的应用通常要求机器人在多方环境中互动。例如，被部署为信息提供者的机器人(在公共信息亭或博物馆中)，与群体或个人一起工作的机器人，以及(在家庭或老年人护理中心)帮助儿童或老年人的机器人。这样的现实背景导致了群体人-机器人交互作为一个新的研究领域的出现。这个项目是关于机器人与群体互动的数据驱动的观点，它将允许机器人系统对个人和群体进行推理。此外，该项目还包括为实现研究人员的长期目标而开展的活动，这些目标是使计算机科学成为一个更加多样化和包容性的领域，并通过人工智能技术增加对科学、技术、工程和数学的参与。例如，项目活动包括与耶鲁皮博迪博物馆合作，帮助主要来自低收入社区的高中生和大学生接受教育，让他们参与研究，并让普通公众参与人工智能和机器人技术。这些协同活动为研究群体人与机器人的相互作用提供了新的机会，从而补充了研究。为了统一在计算上理解群体HRI的许多问题，该项目包括一个三管齐下的方法：1)通过图抽象表示数据，2)通过图神经网络学习，3)通过自我监督以可扩展的方式收集数据。该项目将通过改进机器人发起和维持互动的方式，以切实的方式展示这一方法。首先，该团队将研究预测用户与公共机器人的参与度的问题，作为考虑社会关系的个人推理的例子。其次，它将研究在HRI中识别交互故障的问题，作为关于群体的整体推理的例子。综上所述，这个项目将展示如何将该方法与机器人决策相结合，验证公共环境中用于数据收集和算法评估的实验协议，并在基本层面上促进我们对HRI的理解，以便机器人能够更好地参与群体社交。该项目由跨部门机器人基础研究计划支持，由工程总监(ENG)和计算机和信息科学与工程(CEISE)共同管理和资助。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

The long-term research goal of this Faculty Early Career Development (CAREER) grant is to advance Human-Robot Interaction (HRI) so that robots can effectively take part in social encounters with multiple users. While social robotics research has traditionally focused on one-on-one interactions, real-world applications typically require that robots interact in multi-party settings. Examples include robots that are deployed as information providers (in public kiosks or in museums), robots that work with group of people or with individuals, and robots that assist children or elderly people (in homes or elderly care centers). Such real-world contexts led to the emergence of group Human-Robot Interaction as a new area of study. This project is for a data-driven perspective for robot-group interactions that will allow robotic systems to reason about individuals and groups. Additionally, this project includes activities in pursuit of the researcher’s long-term goals of making Computer Science a more diverse and inclusive field, and increasing engagement in science, technology, engineering, and mathematics via Artificial Intelligence technologies. For example, project activities include a partnership with the Yale Peabody Museum to help educate high-school and college students from primarily low-income communities, engage them in research, and engage the general public with Artificial Intelligence and Robotics. These synergistic activities complement the research by providing novel opportunities to study group human-robot interactions. To unify many problems in computationally understanding group HRI, this project comprises a three-pronged approach that addresses: 1) data representation via graph abstractions, 2) learning via Graph Neural Networks, and 3) data collection in a scalable manner via self-supervision. The project will demonstrate this approach in a tangible manner by improving how robots initiate and sustain interactions. First, the team will study the problem of forecasting user engagement with a public robot as an example of reasoning about individuals in consideration of social relationships. Second, it will study the problem of identifying interaction breakdowns in HRI as an example of reasoning holistically about groups. Taken together, this project will demonstrate how the approach can be integrated with robot decision making, validate experimental protocols for data collection and algorithm evaluation in public settings, and advance our understanding of HRI at a fundamental level so that robots can better take part in group social encounters.This project is supported by the cross-directorate Foundational Research in Robotics program, jointly managed and funded by the Directorates for Engineering (ENG) and Computer and Information Science and Engineering (CISE).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.