Common error sources of industrial robot manipulators include joint servoing error, imprecise kinematics, mechanical compliance, and transmission error. In this work we present a nested loop iterative learning control (ILC) feedforward structure: an inner loop that compensates for motor dynamics, and an outer loop that corrects the deviation along the path tracked, that features practically efficient implementation. Taking advantage of industrial robot’s speed reduction transmission, single-input-single-output method is demonstrated effective for the nonlinear coupled robot dynamics. Data-based inversion technique that incorporates motion constraint is used for fast inner loop convergence. The outer loop utilizes inverse Jacobian matrix for joint reference modification. For nonlinear static friction that is difficult to be compensated for with only joint command, notch filtering is utilized in the learning process to avoid exciting vibration inherently exists in the robot. The proposed nested loop ILC requires only the nominal kinematic parameters from the robot manufacturer, and can be readily implemented without modifying the existing robot controllers. The effectiveness of the proposed method is experimentally verified on a six degree-of-freedom robot manipulator.

工业机器人机械手的常见错误源包括关节伺服错误，运动学不精确，机械顺应性和传输错误。在这项工作中，我们提出了一个嵌套循环迭代学习控制（ILC）前馈结构：一个补偿电机动态的内循环，以及一个校正沿跟踪路径的偏差的外循环，该循环具有实用的实现效率。利用工业机器人的减速传动，证明了单输入单输出方法对于非线性耦合机器人动力学有效。结合了运动约束的基于数据的反演技术用于快速内循环收敛。外循环利用逆雅可比矩阵进行联合参考修改。对于仅靠关节命令很难补偿的非线性静摩擦，在学习过程中利用陷波滤波来避免机器人固有存在的激振。提议的嵌套回路ILC仅需要机器人制造商的标称运动学参数，并且可以在不修改现有机器人控制器的情况下轻松实现。该方法的有效性在六自由度机器人操纵器上进行了实验验证。