NRI: FND: Probabilistic Hypothesis-Driven Adaptive Human-Robot Teams

考虑到机器人技术的成熟以及对每个人（人类或机器人）单独和协作任务的能力，人类和机器人团队的合作有可能改变搜索和救援行动。示例应用包括在地震或火灾后定位幸存者或气体泄漏，其中第一响应者的安全访问受到限制，以及定位城市环境中的化学/生物释放。灾难响应的例子提供了一个广泛的激励挑战，考虑到人类的潜在危险情况，被困的幸存者，快速变化的条件和通信的不确定性。开发真正高效和高性能的人机团队的关键挑战是使团队能够智能地推理和共同行动。 该研究项目将通过开发基于假设的方法来解决这一挑战，以便在人类和机器人之间进行通信，使他们能够对环境进行集体感知，并开发子团队和任务来解决搜索问题。更广泛的教育影响包括让本科生和高中生与研究团队合作进行实验和数据收集。成果包括为社区研究人员以及机器人和控制课程提供开源算法和数据日志;通过出版物、会议、研讨会和NRI会议进行传播;包括本科生和高中生以及在人机合作的跨学科领域合作的多样性项目。该研究项目的目标是为人类开发基础理论和验证算法，机器人团队可以独特地在复杂和不断变化的环境中操作并适应，特别是随着环境/任务的知识随着时间的推移而发展。技术方法使用概率假设公式作为基础，制定过程推理和团队形成的问题。形式化建模方法将人类的自然语言与感知和团队任务的假设联系起来，从而使人类和机器人能够有效地进行协作。机器人将使用物理和数据驱动模型来评估假设的信息内容，以捕获过程和传感。重要的是，推理和团队合作都将随着复杂过程的发展而发展。基于假设的方法和团队适应将在一系列的人机搜索实验中得到验证，并将通过大规模模拟验证缩放。该方法还使感知和规划能够随着场景的发展而发展。该项目旨在通过人机信息交流促进机器人合作;合作机器人团队的可扩展性，其中团队本身随着时间的推移而适应信息收集和知识形成;并展示了智能系统作为人类的物理体现的作用-该奖项反映了NSF的法定使命，并通过使用基金会的学术价值和更广泛的影响审查标准。

Teams of humans and robots working together have the potential to transform search and rescue operations, given the maturation of robotic technology and the ability to task each (human or robot) individually and collaboratively. Example applications include locating survivors or gas leaks after earthquakes or fires, where the safe access for first responders is limited, and locating a chemical/biological release in an urban environment. The disaster response example offers a broad motivational challenge, given the potentially hazardous situations for humans, trapped survivors, fast changing conditions, and communication uncertainties. The key challenge in developing truly efficient and high performing human-robot teams is enabling the team to intelligently reason and act together. This research project will address this challenge by developing hypothesis-based methods to communicate between humans and robots to enable their collective perception about the environment, and to develop sub-teams and tasks to address the search problem as it evolves. Broader educational impacts include having undergraduate and high school students collaborate with the research team to perform experiments and data collection. Outcomes include open source algorithms and data logs for researchers in the community, and for robotics and controls classes; dissemination through publications, conferences, workshops and NRI meetings; and the inclusion of undergrad and high school students and diversity programs collaborating in the interdisciplinary area of human-robot teaming.The goal of this research project is to develop foundational theory and validated algorithms for human-robot teams which can uniquely operate in, and adapt to, a complex and changing environment, particularly as knowledge of the environment/tasks evolves over time. The technical approach uses a probabilistic hypothesis formulation as a basis to formulate both the Process Inference and Team Forming problems. Formal modeling methods will connect human's natural language to hypotheses of the perception and teaming tasks, thereby allowing humans and robots to communicate efficiently and collaboratively. Hypotheses will be evaluated by the robots for information content using physical and data driven models to capture the processes and sensing. Importantly, both the inference and teaming will evolve as the complex processes evolve. The hypothesis-based approaches and team adaptation will be validated in a series of human-robot search experiments, and scaling will be validated via large scale simulations. The approach also enables the perception and planning to evolve as the scene evolves. This project aims to advance cooperative robot collaboration via human-robot information exchange; the scalability of cooperative robot teams where the team itself adapts over time as information is collected and knowledge is formed; and demonstrate the role for physical embodiment of intelligent systems as human-robot teams address complex applicable models.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.