Shared Decision Making and Progressive Automation for Manufacturing Assembly

加拿大制造业正面临着对创新的迫切需求，以提高制造业的生产率、准确性、灵活性和安全性。该研究项目旨在开发下一代机器人技术，以实现制造业装配操作的自动化，为加拿大企业提高生产率、准确性、灵活性、可重构性和易用性。我们的方法背后的一个关键思想是开发渐进自动化工具，从现有的（手动）装配过程开始。这种方法将使公司能够逐步实现装配过程的自动化，降低资本成本并增加灵活性，甚至可以实现快速变化的装配过程的自动化。这将通过五项关键技术的发展来实现：可重构机器人、精确的设施广泛感知和定位、在线任务学习、监测和重新规划以及人机交互。我们将开发可重新配置的机器人，以适应快速变化的生产需求，并且可以很容易地重新配置以处理装配线上的任务变化。将开发一种组合的静态和动态运动捕捉系统，以实现对装配部件、工具、机器人平台和人类操作员的跟踪。拟议的系统将与人类操作员密切合作，以促进任务学习、操作员指导和共享自主决策。当机器人开始执行（一部分）任务时，操作员的反馈将用于纠正和适应。操作员运动的精确测量也将用于允许系统与操作员同步其动作，并在检测到任务序列偏差时向操作员提供反馈并修改计划。该方法将通过使机器人的反馈和控制易于访问、理解和有用，促进人和机器人团队成员之间的沟通，并最大限度地利用人类知识来完成团队任务，从而实现人机团队的无缝操作。

The Canadian manufacturing industry is facing a critical need for innovation to improve the productivity, accuracy, flexibility and safety in manufacturing. This research project aims to develop the next generation of robotic technologies to enable automation of assembly operations in manufacturing, driving improvements in productivity, accuracy, flexibility, reconfigurability and ease of use for Canadian businesses. A key idea behind our approach is the development of tools for progressive automation, starting from existing (manual) assembly processes. This approach will enable companies to incrementally automate the assembly process, reducing capital costs and increasing flexibility, and enabling automation of even rapidly changing assembly processes. This will be accomplished through the development of five key technologies: reconfigurable robotics, accurate facility wide perception and localization, online task learning, monitoring and replanning, and human-machine interaction.We will develop reconfigurable robots that can be adapted to rapidly changing production requirements, and which can be easily reconfigured to handle task changes on the assembly line. A combined static and dynamic motion capture system will be developed to enable tracking of assembly components, tools, robotic platforms and human operators. The proposed system will operate in close collaboration with human operators to facilitate task learning, operator guidance and shared-autonomy decision making. Operator feedback will be used for correction and adaptation as the robot starts to perform (a subset of) the tasks. The accurate measurement of operator movement will also be used to allow the system to synchronize its actions to the operator, and provide feedback to the operator and modify plans when a task sequence deviation is detected. The approach will enable seamless operation of the human-robot team by making feedback from and control of the robot easily accessible, understandable and useful, facilitating communication between human and robot team members, and maximally leveraging human knowledge to accomplish team tasks.