机器人视觉伺服中CMAC学习控制系统研究

根据小脑模型关联控制器（CMAC）收敛速度快，适于实时控制系统的特点，设计了一种基于CMAC学习控制方法的机器人视觉伺服系统。在该系统中，CMAC被用作前馈视觉控制器对常规反馈控制器进行补偿。所提出的CMAC控制器替代图像雅可比矩阵来获得目标图像特征和机器人关节运动之间2D/3D变换关系，通过其在线学习，可以使系统对摄像机标定误差不敏感，从而提高系统的鲁棒性。实验证明了所设计控制系统的有效性。

Research on CMAC study control strategy for robotic visual servo system

Cerebella Model Articulation Controller（CMAC） neural network has fast learning speed and can be used very well for real time system.A study control strategy for robotic visual servo system is introduced.In this strategy, CMAC is used as feedforward visual servo controller to compensate a general feedback controller.The proposed CMAC controller approximates the image Jacobin matrix that maps the 2D/3D relationships between image features and robotic joints movements. The on-line learning ability of CMAC can make the proposed system not sensitive to the camera calibrated errors and has strong robustness.The results of experiments prove the validity of the presented control system.