HCC: Small: Leadership Emergence for Synchronous Human-Autonomy Teaming

机器越来越自主，能够在很少或没有监督的情况下执行许多复杂的任务。然而，在许多现实世界的应用程序中，他们仍然经常需要与人类一起工作，在这些应用程序中，人类提供各种各样的、冗余的和互补的技能，使系统更具弹性。因此，在无人驾驶汽车和医疗诊断等新兴应用中，能够制造能够与人类良好合作的自主机器将变得越来越重要。该项目的研究方向是让机器像人类一样在大型组织中进行协作，特别是在领导、跟随和与他人同步方面，以完成日常和紧急情况下的关键任务。最终，该项目将增强智能机器与人类合作的能力，以改善商业运作，并以稳健可靠的方式为公众服务。该项目旨在定义领导力和认知同步的涌现性，以便将自主性集成到多代理组织中。项目团队将以内隐领导理论为基础，该理论预测，人类团队中的领导和跟随动态倾向于将领导权授予最有能力和最值得信赖的成员，而不管预先定义的权力结构如何。为了利用这一理论，研究人员将根据技能、规则和基于知识的行为来开发方法，以支持不同任务之间的认知同步，这些行为将人类认知分类并作为信任评估的基础。将开发强化学习方法来选择交互方法和改进交互选择，以促进长期、富有成效的协作。在搜索和救援任务的背景下，这些方法将在计算机模拟和无人驾驶车辆和人类队友的现场实验中进行评估。设计原则和算法将适用于许多涉及人类在循环中的命令和控制的领域，例如灾难响应，空中交通管制和患者流管理。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。

Machines are increasingly autonomous, capable of performing many complex tasks with little or no supervision. However, they still often need to work on teams with humans in many real-world applications, where humans provide diverse, redundant, and complementary skills that make the systems more resilient. Thus, being able to build autonomous machines that can team well with humans will become increasingly important in emerging applications such as driverless cars and medical diagnosis. The research in this project moves toward allowing machines to collaborate as humans do in large organizations, especially in terms of leading, following, and synchronizing with others to complete critical tasks in both routine and emergency situations. Ultimately, the project will enhance the abilities of intelligent machines to work with humans to improve business operations and serve the public in a robust and reliable manner.The project seeks to define the emergent nature of leadership and cognitive synchronization for integrating autonomy into multi-agent organizations. The project team will ground the work in implicit leadership theory, which predicts that leading and following dynamics in human teams tend to result in granting leadership to the most competent and trustworthy member regardless of a pre-defined authority structure. To leverage this theory, the researchers will develop methods to support cognitive synchronization across diverse tasks according to skill-, rule-, and knowledge-based behaviors that classify human cognition and underlie assessments of trust. Reinforcement learning approaches will be developed for selecting interaction methods and improving interaction choices to promote long-term, productive collaboration. These methods will be evaluated in computer simulations and field experiments with unmanned vehicles and human teammates in the context of search and rescue missions. The design principles and algorithms will be applicable to many domains involving command and control with humans in the loop, such as disaster response, air traffic control, and patient flow management.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.