Analysis, design and control of collaborative robots for extended reach dexterous assembly

装配是制造复杂产品所需工作量的重要组成部分。由于组装需要很高的灵巧性和灵活性，因此通常仍由操作员进行。然而，人类操作员受到人体工程学限制，如有限的伸展范围、力量、姿势和其他因素。另一种选择是使用自动化。然而，尽管机器人技术最近取得了进步，但今天的自主机器人无法达到人类操作员特有的速度和可靠性标准。因此，涉及人类和协作机器人的解决方案-支持人类操作员的努力-在最近的文献中进行了调查。在这项研究项目中，建议通过开发一种人工辅助解决方案来逆转这一范式。在这个概念中，组装机器人设备将在人类操作员的全球控制下运行。拟议中的辅助设备将直接参与灵活的组装，从而允许任务超出人类的触角或力量限制。首先，将针对一类有针对性的灵活任务进行组装机制的建模，包括集成被动和/或主动解决方案以及所需的控制算法。然后，将开发在硬件和控制中提供增力的方法。然后将研究两类机器人概念，即具有有限能力的简单机械概念，以及具有高级人类辅助能力的完全主动解决方案。我们将分别为这两种类型的解决方案构建一个原型。原型将被用来证明所提议的方法在制造环境中的有效性。然后，将利用传感和统计分析技术制定战略，以可靠地检测拟议原型的装配任务的完成情况。这包括一个适用于工业应用的机器学习框架。最后，将开发简单的人机界面，以演示所开发的机器人概念的直观编程以及它们在工业应用中的相关性。

Assembly represents a significant portion of the effort required for the manufacturing of complex products. Since assembly requires a high level of dexterity and flexibility, it is still generally performed by human operators. Human operators, however, are limited by ergonomic constraints such as limited reach, forces, posture and other factors. An alternative is the use of automation. However, despite the recent advances in robotics technology, autonomous robotics today cannot meet the standards of speed and reliability that are characteristic of human operators. Therefore, solutions involving humans and collaborative robots - that support the efforts of human operators - have been investigated in the recent literature. In this research project, it is proposed to invert this paradigm by developing a human-assisted solution. In this concept, assembly robotic devices will be designed that function under the global control of human operators. The proposed assistive devices will directly participate in the dexterous assembly, and thus allow tasks to extend beyond the reach or force constraints of humans. First, the modelling of the mechanics of assembly will be pursued for a class of targeted dexterous tasks including the integration of passive and/or active solutions and the control algorithms needed. Then, means of providing force augmentation in hardware and controls will be developed. Two families of robotic concepts will then be investigated, namely a simple mechanical concept with limited capability, and a fully active solution with advanced human-assisted capability. A prototype will be built for each of these two types of solutions. The prototypes will be used to demonstrate the effectiveness of the proposed approach in a context of manufacturing. The development of strategies for the reliable detection of the completion of an assembly task with the proposed prototypes will then be pursued using sensing and statistical analysis techniques. This includes a machine learning framework suitable for an industrial application. Finally, simple Human-Machine interfaces will be developed in order to demonstrate the intuitive programming of the developed concepts of robots and thereby their relevance in industrial applications.**********