NRI: FND: Improving Human-Robot Collaboration on Assembly Tasks by Anticipating Human Actions

人类安全机器人的实现将为人类-机器人团队在复杂的装配任务中的部署提供机会。人类可以执行需要灵活的复杂任务，而机器人可以执行不需要很高灵活性的辅助任务。让人类和机器人在近距离作业，同时利用它们的互补优势，可以显著提高人类的生产力，减轻人类的工作压力。当人类在团队中工作时，团队成员养成流畅的团队行为是很重要的，对于人类-机器人团队也应该如此。这就需要机器人助手适应人类队友的喜好，预测他们的行动，并支持他们执行任务。该奖项支持基础研究，以实现在装配制造任务中与人类队友合作的多个机器人的动作适应。研究结果将有助于在装配制造任务中引入多个工业机器人，并将为美国制造业带来增长机会。将研究与研究生和本科课程相结合，将加强机器人和制造课程，并丰富参与学生的学习经验。推广活动将教育和告知K-12学生在机器人和制造领域的职业机会。要有效地适应多个机器人在人类-机器人杂交细胞中的作用，需要在人类偏好建模、人类行动预测和人类感知任务和运动规划方面取得基本进展。这项研究将调查机器学习算法设计的计算基础，该算法识别人类操作员在规范组装任务中的主导偏好。将开发用于生成指定需要执行的任务的任务计划并将其分配给系统中的各种代理的算法，以及用于使用任务计划线索和工作单元监控来预测下一人工行动的算法。这项研究将探索和描述使用预测动作来适应多个机器人在装配任务中与人类队友交互的动作的方法。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。

The realization of human-safe robots would present opportunities for the deployment of human-robot teams in complex assembly tasks. Humans can perform complex tasks that require dexterity, while robots can perform supporting tasks that do not require a high degree of dexterity. Having humans and robots operate in close proximity, while utilizing their complementary strengths, can significantly enhance human productivity and reduce job stress for humans. When humans work in teams, it is important for the team members to develop fluent team behavior and the same should hold for a human-robot team. This requires robotic assistants to adapt to the preferences of human teammates, anticipate their actions and support them in performing the task. This award supports fundamental research to enable adaptation of the actions of multiple robots collaborating with a human teammate in an assembly manufacturing task. Results from the research will facilitate introduction of multiple industrial robots in assembly manufacturing tasks and will result in growth opportunities for the US manufacturing industry. The integration of the research with graduate and undergraduate courses will enhance robotics and manufacturing curricula and enrich the learning experiences of the participating students. Outreach activities will educate and inform K-12 students about career opportunities in robotics and manufacturing.Effective adaptation of multiple robots in human-robot hybrid cells requires fundamental advances in human preference modeling, human action prediction and human-aware task and motion planning. This research will investigate computational foundations for the design of machine learning algorithms that identify the dominant preferences of human operators in canonical assembly tasks. Algorithms will be developed for generating task plans that specify tasks that need to be performed and assigning them to various agents in the system and for predicting the next human action using task plan cues and work cell monitoring. This research will explore and characterize methods for using the predicted actions to adapt the actions of multiple robots interacting with a human teammate on an assembly task.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.