CAREER: Uncovering Structure in Human-Robot Systems for Trajectory Prediction and Crowd Navigation

智能机器人和自主系统在我们的日常生活中迅速变得司空见惯。然而，只有当底层算法能够处理人类面临的独特挑战时，自主性的理想影响才能实现。为了设计安全、值得信赖的自主系统，需要改变智能系统交互、影响和预测人类代理的方式。这个学院早期职业发展(Career)项目将专注于了解人类和移动机器人如何能够以及应该如何互动。首先，将考虑通过结构化表示来预测拥挤空间中的人的轨迹。其次，将考虑使用这些预测进行智能决策的机器人策略。其目的是平衡效率和安全性，即使在人类行为反复无常和传感器不确定的情况下，也能确保可靠的性能。这些方法将在真实机器人上进行评估，动机是高影响的问题领域，包括农业机器人(目前面临劳动力危机，导致对机器人的需求增加)；协作制造(看到合作机器人的受欢迎程度上升)；以及交通(在交通领域，行为预测和交互仍然是自主的关键挑战之一)。该项目将通过开发PrairieLearn的机器人课程、伊利诺伊大学香槟分校(UIUC)开发的在线问题驱动学习系统以及K-12外联活动来支持机器人教育，以激发人们对STEM和机器人技术的兴趣。在线教程和短期课程也将有助于研究与教育的整合和向产业的过渡。机器人和自动化研究的十年正到达一个关键转折点：理论和全面部署之间的差距开始缩小。然而，只有当底层算法能够成功地考虑人类行为时，自主性的理想效果才能实现。为了设计可信赖的系统，需要改变智能系统交互、影响和预测人类的方式。这个项目将研究人类和移动机器人之间的互动，旨在揭示这种互动中的潜在结构，并使流体机器人能够导航。将研究轨迹预测和人群导航。首先，交互图将作为一种形式表示引入，它捕获代理之间的耦合，并通过因子分解允许易管理性和计算效率。这个框架允许对交互类型进行建模，以捕获代理之间的可变和动态关系。其次，这种表示和预测洞察力将被结合到稳健的导航中，即使在存在不确定性的情况下也能提供安全。改善移动机器人的交互将影响许多不同的应用领域，包括农业机器人、协同制造和交通运输。该项目将通过开发PrairieLearn中的机器人课程来支持机器人教育，这是一种由UIUC开发的在线问题驱动的学习系统，以及为有兴趣了解这一领域研究的行业合作伙伴提供的教程。该项目由跨部门机器人基础研究计划支持，该计划由工程学指导委员会(ENG)和计算机与信息科学与工程指导委员会(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。

Intelligent robots and autonomous systems are quickly becoming commonplace in our daily lives. However, the desirable impacts of autonomy are only achievable if the underlying algorithms can handle the unique challenges that humans present. To design safe, trustworthy autonomous systems, there is a need to transform how intelligent systems interact, influence, and predict human agents. This Faculty Early Career Development (CAREER) project will focus on the understanding how humans and mobile robots can and should interact. First, predicting human trajectories in crowded spaces through structured representations will be considered. Second, robot strategies for using these predictions for intelligent decision-making will be considered. The aim is to balance efficiency and safety, guaranteeing reliable performance even in the presence of erratic human behavior and sensor uncertainty. The approaches will be evaluated on real-world robots, motivated by high-impact problem domains, including agricultural robots (which is currently facing a labor crisis, resulting in an increased demand for robotics); collaborative manufacturing (which is seeing a rise in popularity of co-robots); and transportation (where behavior prediction and interaction remains one of key challenges for autonomy). This project will support robotics education through the development of robotics coursework in PrairieLearn, an online problem-driven learning system developed at University of Illinois Urbana-Champaign (UIUC), and K-12 outreach to spur interest in STEM and robotics. On-line tutorials and short courses will also aid in both integration of research with education and transition to industry.Decades of robotics and automation research is reaching a critical turning point: the gap between theory and full-scale deployment is beginning to close. However, the desirable impacts of autonomy are only achievable if the underlying algorithms can successfully consider human behavior. To design trustworthy systems, there is a need to transform how intelligent systems interact, influence, and predict humans. This project will examine the interaction between humans and mobile robots, aiming to uncover the underlying structure within this interaction and enable fluid robot navigation. Trajectory prediction and crowd navigation will be examined. First, interaction graphs will be introduced as a formal representation that captures the coupling between agents and allows for tractability and computational efficiency through factorizations. This framework allows modeling types of interactions that capture variable and dynamic relationships between agents. Second, this representation and prediction insight will be combined into robust navigation, providing safety even in the presence of uncertainty. Improving mobile robot interaction will impact many different application sectors, including agricultural robots, collaborative manufacturing, and transportation. This project will support robotics education through the development of robotics coursework in PrairieLearn, an online problem-driven learning system developed at UIUC, and tutorials for industry partners interested in learning about research in this area. 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.