基于多机器人协调的大型舱段轴孔精密装配控制方法研究

Large bay section assembly is a key technology in the manufacturing of national defense equipment. However, the traditional assembly method cannot fulfill the precise assembly of large bay section automatically. In this project, high precision pole-in-hole assembly of large bay section is chosen as the objective. The multi-robot coordination control method is developed and a prototype is designed. This project focuses on the key technical breakthrough in the following three aspects: (1) The rapid and accurate measurement of pose information of large bay section. Combining the deep neural network and ROI, position marker can be roughly located in the global image. Then the structured light vision is used to achieve the precise measurement of the position marker in the local range. (2) The precise measurement of the contact force of large bay section pole-in-hole assembly. The distributed force sensor model can be obtained by the dynamic analysis of the bay section and supporting mechanism, while the signal error from the force sensor is analyzed and the sensor is calibrated accordingly. Thus, the precise measurement of the contact force in assembly process is realized. (3) High precision control method of large bay section pole-in-hole assembly. A hierarchical intelligent control model is adopted. Then an intelligent control method based on force and position information feedback and reinforcement learning is presented, which reduces the system’s dependence on the mechanical model and the sensor precision. Therefore, the high precision pole-in-hole assembly of large bay section is achieved.

大型舱段装配是国防装备制造的关键技术，现有的方法无法自动完成大型舱段轴孔精密装配。本项目以大型舱段轴孔精密装配为目标，研究多机器人协调控制方法，研制原理样机，并重点突破以下三方面关键技术：(1) 大型舱段位姿信息快速准确测量。采用深度神经网络与ROI相结合在全局图像中实现标志点的粗定位，利用结构光视觉在局部范围内完成标志点的精确测量。(2) 大型舱段轴孔装配接触力的精密测量。通过对舱段和支撑机构的动力学分析，得到分布式力传感器理论模型，并对力传感器信号进行误差分析及校准，实现装配接触力的精密测量。(3) 大型舱段轴孔高精度装配控制方法。采用分层递阶的智能控制模式，提出基于力位信息反馈和强化学习的控制方法，降低对于力学模型和传感器精度的依赖，实现大型舱段轴孔高精度装配。

舱段对接工艺是航空航天和军工制造中常见的生产过程，对接装配技术直接影响最终产品的生产周期、一致性以及性能,而目前我国的舱段装配方式智能化和自动化程度较低。本课题研究多机器人协同的舱段精密自主装配方法。首先，完成了多机器人协同的装配系统设计，并提出一种灵巧工作空间的快速求解方法。在舱段位姿测量方面，提出了一种基于全局立体视觉和局部手眼结构光视觉的大型工件位姿测量系统和方法。在大型舱段轴孔装配接触力的测量方面，提出一种基于旋量几何的动力学建模方法，完成了负载分配分析以及舱段接触力辨识。在大型舱段轴孔精密装配的控制方法方面，提出了一种三阶段力反馈柔顺控装配方法，实现了舱段接触到轴孔插入再到定位结构精确配合的全流程柔顺作业。另外在系统标定方面，提出了一种基于视觉的舱段坐标系标定与标定位形规划方法，一种线结构光手眼系统的内外参高效标定流程，以及一种不依赖标定物的结构光手眼自标定方法。通过以上研究及所搭建原理样机，实现了高效、自主、柔顺的舱段精密装配。