摘要：

提出一种机器人捕捉运动目标的动力学视觉伺服方法。基于位置阻抗控制器，通过双目立体视觉检测并跟踪运动目标的位置，结合CMAC网络，采用以视觉阻抗的二次型为训练目标的学习型视觉阻抗控制器，用于克服控制器对系统结构参数变化适应性差的缺点，并对阻抗参数进行优化。实验结果表明，该视觉伺服方法在机器人捕捉运动目标时具有良好的动力学特性和轨迹控制效果。

关键词: 运动目标捕捉, 视觉阻抗控制, CMAC网络, Kalman滤波器

Abstract:

This paper proposes a dynamics visual servo method for robot’s grasping moving object. The motion object target position is detected and tracked by dual stereo vision. With CMAC network, a learning visual impedance controller taking the quadratic of vision impedance as training target is adopted. It promotes the adaptability of position controller of the system structure parameters, so the impedance parameters are optimized. Experimental results show that the dynamics characteristics are improved and a flexible trajectory control is obtained with the method.

Key words: moving object grasping, vision impedance control, CMAC network, Kalman filter

中图分类号: 

• TP391.41