CAREER: Modeling the Dynamic Interplay of Control, Planning and Perceptual Functions in Agile Human Guidance

这项教师早期职业发展（Career）计划的研究目标是研究和模拟空间控制、感知和规划功能在人类引导技能中的相互作用。有充分的经验证据支持有经验的飞行员或操作员在他们的行为中表现出模式的观察。这些模式很重要，因为它们是人类用来组织和实现敏捷指导行为的机制的表现。用小型直升机进行的制导任务将用于生成结合车辆运动、飞行员控制和视觉注视信息的实验数据。研究分为三个方面：1)采用数据驱动的方法，结合动力学、控制、识别和数据聚类方法，确定模式产生的原理；2)分析模式之间的转换及其动态组成，确定底层模型结构；最后3)整合所获得的知识，形成描述控制、感知和规划功能之间相互作用的交互模型。如果成功，这项研究将使远程操作系统能够利用操作员天生的引导功能来减少工作量并提高安全性。示例应用包括搜索和救援直升机或远程手术系统。这些系统将引导操作员的视觉注意力到对性能有重要意义的线索上，反过来，解码视觉注意力来预测未来的控制动作。拟议的研究活动将促进控制工程和认知科学领域之间的互动。将利用跨学科的机会：1)创建刺激的本科和研究生课程；2)鼓励代表性不足的学生参与研究环境；3)为高中和小学设计外展活动。研究结果将通过一个向广大公众开放的网站和通过旨在使控制界和人为因素界接触的会议讲习班加以传播。

The research objective of this Faculty Early Career Development (CAREER) Program grant is to study and model the interplay of spatial control, perception and planning functions in human guidance skills. Ample empirical evidence exists supporting the observation that experienced pilots or operators display patterns in their behavior. These patterns are significant because they are manifestations of the mechanisms used by humans to organize and implement agile guidance behaviors. Guidance tasks conducted with small-scale helicopters will be used to generate experimental data that combine vehicle motion, pilot control and visual gaze information. The research is organized in three threads: 1) Determine the principles of patterns' emergence using a data-driven approach combining dynamics, control, identification and data-clustering methods; 2) Analyze the transition across patterns and their dynamic makeup to determine the underlying model structures; and finally 3) Integrate the gained knowledge to form an interaction model that describes the interplay between control, perceptual and planning functions. If successful this research will enable tele-operation systems that take advantage of the operator's innate guidance functions to reduce workload and increase safety. Example applications include search and rescue helicopters or tele-surgery systems. These systems will work by directing the operator's visual attention to cues that are significant to performance and conversely, decode visual attention to predict future control actions. The proposed research activities will promote interactions between the fields of control engineering and cognitive sciences. Cross-disciplinary opportunities will be leveraged to: 1) Create stimulating undergraduate and graduate courses; 2) Encourage the involvement of underrepresented students in the research environment; 3) Design outreach activities for high-schools and elementary schools. The results will be disseminated through a website accessible to the broader public and through conference workshops designed to reach across the control and human factor communities.