CAREER: Making Robots More Cooperative Agents: Controlling Costs of Coordination Through Graph-Based Models of Joint Activity

智能机器人的部署保证了在灾难和紧急响应、地面机动性、制造、航空和空间操作等领域提高安全性、生产力和能力。良好的人-机器人协作是实现这些承诺的关键。该项目开发了新的建模技术，用于分析和设计人-机器人团队中的协作行为。协作行为需要彼此适应和沟通。协调和沟通会产生认知和时间成本。在人与机器人的协作中，这些成本可能很高，因为与自主代理的协调通常比与其他人类的协作更繁琐和耗时。该项目开发的模型将确定人类-机器人系统中协调成本的原因和影响。基于这些模型，该项目开发了管理协调成本的技术，以避免操作员超负荷工作。改进的成本管理将导致更强大和更具弹性的人类-机器人操作，更广泛地采用智能机器人技术，并实现其承诺。该项目将研究与教育和外联活动相结合，以培训未来的工作人员进行系统思维和跨学科解决问题的技能。这些技能将为未来的工程师、研究人员和科学家创建集成的解决方案做好准备，以应对跨越技术、人类、生态、经济和政策维度的复杂挑战。该项目将认知科学和社会科学的理论与图论和基于代理的建模技术相结合，开发了一种通用的形式化方法，用于表示和分析人类-机器人系统中的关节活动。该框架允许对管理人类和机器人之间的相互依赖所需的团队工作进行客观和动态的分析。在该模型的基础上，研究开发了动态适应和控制协调成本的技术，以改善协作和避免失误。这项工作将在灾害应对和空间行动中得到验证。该项目解决了三个基本的研究挑战：首先，它确定了人-机器人系统组织、合作能力的不对称性以及与机器人协调的认知和时间成本之间的关系。其次，确定了人-机器人系统中动态调节协调成本的控制策略。第三，它展示了使用图论度量和算法将联合活动的理论概念转化为可操作的指导，以使机器人在动态环境中更具协作性。这些发现将为机器人需要被赋予哪些能力，以使它们在上下文中成为有用的合作代理提供深刻的见解。这些见解将告诉我们应该如何部署机器人功能，以提高复杂操作的稳健性和弹性。这一奖项反映了NSF的法定使命，并通过使用基金会的智力优点和更广泛的影响审查标准进行评估，被认为值得支持。

The deployment of smart robots promises increased safety, productivity, and capability in domains such as disaster and emergency response, ground mobility, manufacturing, aviation, and space operations. Good human-robot collaboration is key to realizing these promises. This project develops novel modeling techniques for analyzing and designing collaborative behavior in human-robot teams. Collaborative behavior requires adjusting to and communicating with each other. Coordination and communication incur cognitive and temporal costs. In human-robot collaboration these costs can be high, as coordination with autonomous agents generally is more taxing and time-consuming than collaboration with other humans. The models developed in this project will identify the causes and effects of coordination costs in human-robot systems. Based on these models, the project develops techniques for managing coordination costs to avoid overloading human operators. Improved cost management will lead to more robust and resilient human-robot operations, broader adoption of smart robotic technologies, and realization of their promise. The project integrates the research with education and outreach activities to train the future workforce in systems thinking and interdisciplinary problem-solving skills. These skills will ready future engineers, researchers, and scientists to create integrated solutions to address complex challenges that span technological, human, ecological, economic, and policy dimensions.This project develops a generalizable formalization for representing and analyzing joint activity in human-robot systems by combining theories from cognitive and social sciences with techniques from graph theory and agent-based modeling. This framework allows objective and dynamic analysis of the teamwork required to manage interdependencies between humans and robots. Based on the model, the research develops techniques for dynamically adapting and controlling coordination costs to improve collaboration and avoid lapses. The work will be validated in disaster response and space operations. The project addresses three fundamental research challenges: First, it determines the relation between a human-robot system organization, asymmetries in cooperative competencies, and cognitive and temporal costs of coordinating with robots. Second, it identifies control strategies for dynamically regulating coordination costs in human-robot systems. Third, it demonstrates the use of graph-theoretical metrics and algorithms to translate theoretical concepts of joint activity into actionable guidance for making robots more cooperative agents in dynamic environments. Findings will provide deep insight into what capabilities robots need to be endowed with to make them useful cooperative agents in context. These insights will tell us how robotic functionality should be deployed to improve the robustness and resilience of complex operations.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.