基于刚柔耦合模型及多重采样空间的轻量化机械臂标定方法研究

Industrial robots are widely used in assembling, logistics, national defense, etc., and have significant engineering applicant values. Its light-weight design meets the requirement of high speed and energy conservation, but increases the flexibility of robot arm. This makes the change regulations and coupling relationship of geometric error and flexible error more complicated. The influence of these errors can be significantly decreased by kinematic calibration, and the positioning accuracy of robot arm is then greatly improved. Traditional calibration method considered geometric error as constant and describe it separately with flexible error. This kind of characterization would bring low-precision model, poor identifiability and undesirable calibration result, when it is used for lightweight robot arm. This project takes the lightweight robot arm as research object. Firstly, the change regulations of geometric error and flexible error are deeply explored to establish rigid-flexible coupling error function which describes geometric error and flexible error uniformly. Then, a measuring method of robot arm gesture based on multiple sampling spaces is proposed to improve the identifiability of rigid-flexible coupling error. Also, the measuring trajectory is optimized within multiple constraints to shorten the time of sampling and reduce the dynamic interference. Lastly, a stepwise identification method based on matrix priority is proposed, which optimizes the identifiability of rigid-flexible coupling error once more. The contribution of this project could provide new theory and engineering applicant method for kinematic calibration technology of lightweight robot arm and other kind of multi-body systems.

工业机器人广泛应用于装配、物流及国防科技等领域，其轻量化设计可满足高速、节能需求，但使得关节柔性明显增强，几何误差和柔度误差变化特征及耦合关系非常复杂。运动学标定可减少此类误差影响，显著提高机械臂绝对定位精度，但传统标定方法将几何误差视为常量且与柔度误差分别表征，面向轻量化机械臂时将导致误差模型精度不足，可辨识性明显降低，无法达到满意标定效果。本项目以轻量化机械臂为对象，深入研究并提炼几何误差和柔度误差共性变化规律，构造刚柔耦合误差函数，实现多类型误差统一表征；探索采样空间及点阵结构的优化途径，并在多约束条件下生成位姿数据样本，进而提出基于多重采样空间及点阵的位姿测量方法，提升采样性能及可辨识性；研究基于优先级的误差参数渐近式辨识方法，实现可辨识性二次优化，并构建完整的运动学标定系统。本项目研究成果能够为轻量化机械臂以及其他刚柔耦合多体系统的运动学标定技术提供新的思路和有益探索。

工业机器人广泛应用于装配、物流及国防科技等领域，其轻量化设计可满足高速、节能需求，但使得关节柔性明显增强，几何误差和柔度误差变化特征及耦合关系非常复杂。运动学标定可减少此类误差影响，显著提高机械臂绝对定位精度，但传统标定方法将几何误差视为常量且与柔度误差分别表征，面向轻量化机械臂时将导致误差模型精度不足，可辨识性明显降低，无法达到满意标定效果。本项目以轻量化机械臂为对象，深入研究并提炼几何误差和柔度误差共性变化规律，构造刚柔耦合误差函数，实现多类型误差统一表征；提出基于多重采样空间及点阵的位姿测量方法，提升采样性能及可辨识性；研究基于优先级的误差参数渐近式辨识方法，实现可辨识性二次优化。本项目研究成果能够为轻量化机械臂以及其他刚柔耦合多体系统的运动学标定技术提供新的思路和有益探索。.项目通过对上述内容的研究和探索，得到了一系列基于多重采样空间的误差参数渐近辨识方法，能够分别用于辨识纯刚性几何误差、含有柔性成分的几何误差、刚柔耦合运动学误差，并将这些成果由传统的六自由度机械臂拓展至柔顺型机械臂以及冗余机械臂，分别提供了辨识模型的建立方法。虽然建立多重采样空间对辨识模型的条件数和观测指数起到极大提升作用，但还有进一步提升空间，因此项目在此基础上还针对辨识模型的稳定性提出了一种矩阵平衡方法，进一步提高了误差参数辨识精度。