CAREER: Cooperative Motion Planning for Human-Operated Robots

该提案概述了一项研究和教育计划，以推进机器人与人类操作员合作的决策技术。由于人类在视觉、创造力和适应性方面远远超过最先进的机器人，人们对以人为中心的机器人方法的兴趣迅速增长：将人类的优势与机器人的卓越精度和可重复性相结合。然而，我们现有的运动规划工具，虽然在计算复杂自主任务的运动方面功能强大，但不适合以人为中心的应用程序，这些应用程序需要响应性和自然运动。该提案假设一种新的协同运动规划范式将支持人类操作机器人（如智能车辆、远程手术系统、搜索和救援机器人和家用机器人）在直观性和任务性能方面的重大进步。这一假设在一项教育计划中得到了回应，该计划旨在培养具有跨学科优势的工程师，从而在机器人技术和社会层面之间架起桥梁。最初对新手操作人员进行的人类受试者研究表明，PI的协同运动规划算法可以显著减少任务完成时间和混乱环境中的碰撞率。拟议的工作将沿着这条研究路线进行进一步的调查，以1)确定协作规划者的特征-例如最优性，响应性和完整性-产生有效的人类操作员系统，无论是在客观性能指标还是主观偏好方面，2)设计在计算资源和通信约束下优化协作指标的规划者。3)增强规划人员协助操作员完成复杂操作任务的能力。在本研究中开发的规划器和通过用户研究获得的丰富数据集将作为资源，帮助人机交互（HRI）研究人员设计安全和社会可接受的机器人行为。此外，协同运动规划方面的进步可能会产生长期的社会和经济影响，使机器人在驾驶辅助系统、太空探索、医学、家用机器人、制造业和建筑业中的新应用成为可能。在一系列活动中，研究与教育相结合，包括计算机科学课程开发，关于优化和机器学习的新研究生课程的开发，以及用于机器人教育的新软件库。新的运动规划、行为识别和人力资源调查模块将被纳入人工智能和机器人课程。每年夏天都会申请一名REU，并将与非洲裔美国计算机研究人员促进联盟（A4RC）合作，从少数族裔服务机构招募REU。一名或多名IU本科生将参与研究，并根据计算机本科生研究机会（UROC）计划进行指导，少数民族和女性学生优先考虑。

This proposal outlines a research and educational plan to advance decision-making techniques for robots that cooperate with human operators. Because humans far exceed the abilities of state-of-the-art robots in vision, creativity, and adaptability, interest is rapidly growing in a human-centered approach to robotics: combining the strengths of humans with the superior precision and repeatability of robots. And yet, our available motion planning tools, while powerful at computing motions for complex autonomous tasks, are poorly suited for human-centered applications that demand responsive and natural motions. This proposal hypothesizes that a new cooperative motion planning paradigm will support major advances in intuitiveness and task performance of human-operated robots such as intelligent vehicles, tele-surgery systems, search and-rescue robots, and household robots. This hypothesis is echoed in an educational plan that aims to train engineers with cross-disciplinary strengths that bridge both the technical and social dimensions of robotics. Initial human subjects studies on novice operators with the PI's cooperative motion planning algorithms suggest that the technique leads to dramatic reductions in task completion time and collision rate in cluttered environments. The proposed work will conduct further investigations along this line of research to 1) identify characteristics of cooperative planners - such as optimality, responsiveness, and completeness - that yield effective human-operator systems, both in terms of objective performance metrics and subjective preferences, 2) to design planners that optimize cooperativity metrics under computational resource and communication constraints, and 3) to enhance the capabilities of such planners to assist operators in complex manipulation tasks.The planners developed in this research and the rich datasets acquired via user studies will serve as resources to help human-robot interaction (HRI) researchers design safe and socially acceptable robot behaviors. Moreover, advances in cooperative motion planning may have long-term social and economic impact by enabling new applications of robotics in driver assist systems, space exploration, medicine, household robotics, manufacturing, and construction. Research is integrated with education in a range of activities that include CS curriculum development, development of a new graduate course on optimization and machine learning, and in new software libraries for robotics education. New modules on motion planning, behavior recognition, and HRI will be incorporated in AI and robotics courses. An REU is requested for each summer of the grant and will be recruited from a minority-serving institution in cooperation with the Alliance for the Advancement of African-American Researchers in Computing (A4RC). One or more IU undergraduates will be involved in research and mentored according to the Undergraduate Research Opportunities in Computing (UROC) program, with preference given to minority and women students.