CAREER: Formal Tools For Analysis and Design of Collaborative Hybrid Systems

本学院早期职业发展（Career）计划奖的研究目标是将可观察性方法、可达性分析和混合控制理论中的控制器综合应用于人类自动化系统的分析和设计。将创建方法和工具来验证和设计协作混合系统。将开发一个正式框架，将以用户为中心的设计的定性人为因素指导方针和人为输入和用户界面的数学模型结合到协作混合系统的数学严格框架中。本研究具体涉及：1)优先，动态驱动的用户界面设计，2)协同控制综合，3)安全验证与人在循环中，以及4)可定制的显示。这些技术将用于1)评估多模态用户界面中的信息内容（建模为混合系统的输出地图），2)识别状态空间中具有潜在“战斗”的区域。在人与自动化之间，3)将状态配置与多个安全级别相关联，4)在不完全信息下评估用户界面的有效性，这些信息表示不准确的心智模型或不同级别的用户培训。如果成功，这项研究的结果将有助于在构建、测试和部署安全关键、高风险或昂贵的协作系统之前识别问题（例如自动化意外、人类操作员与自动化之间的冲突、用户界面中包含的信息不足或过多，以及错误或频繁的警报）。本文提出的研究和教育计划将有助于人机交互领域的形式化方法，以及控制理论领域对决策支持辅助设计的关注。开发的理论、代码和数据将及时传播。研究生和本科生将受益于包含主动学习技巧的新课程和辞职课程，以及通过指导和参与拟议的研究。

The research objective of this Faculty Early Career Development (CAREER) Program award is to apply observability methods and reachability analysis and controller synthesis from hybrid control theory to the analysis and design of human-automation systems. Methods and tools will be created to validate and design collaborative hybrid systems. A formal framework will be developed that incorporates both qualitative human factors guidelines for user-centered design, and mathematical models of human input and user interfaces, into a mathematically rigorous framework for collaborative hybrid systems. This research specifically addresses: 1) prioritized, dynamics-driven user-interface design, 2) collaborative control synthesis, 3) verification of safety with a human in the loop, and 4) customizable displays. These techniques will be used to 1) assess information content in multi-modal user interfaces (modeled as output maps of hybrid systems), 2) identify regions in the state-space with potential for ?fighting? between the human and the automation, 3) associate state configurations with multiple levels of safety, and 4) assess efficacy of user-interfaces under imperfect information that represent inaccurate mental models or various levels of user training.If successful, the results of this research could help identify problems (such as automation surprises, fighting between the human operator and the automation, inadequate or excessive information contained in the user interface, and false or frequent alarms) in safety-critical, high-risk, or expensive collaborative systems before they are built, tested, and deployed. The proposed research and education plan will contribute to the formal methods in human computer interaction community, and to the control theory communities focused on the design of decision support aids. Developed theory, code, and data will be disseminated in a timely manner. Graduate and undergraduate students will benefit through new and resigned courses that incorporate active learning techniques, as well as through mentoring and participation in the proposed research.