【摘要】 为实现水轮机修复专用机器人的位姿控制,依据神经网络可以逼近任何非线性函数这一特性,采用融合遗传算法的神经网络建立机器人逆运动学模型,针对机器人的结构特点,改进了神经网络的输出训练样本,设计了用遗传算法学习神经网络权系数的软件实现方法,并提出了应用轨迹插补算法取得轨迹中间点位姿.运行结果表明,该方法可大大提高机器人的位姿控制准确度,并能快速地实现位姿控制.更多还原

【Abstract】 In order to realize the position and pose control of the special robot repairing hydraulic turbines,according to neural networks capable of approximating any nonlinear function, a kind of neural networks mixed with genetic algorithm is applied to build the inverse kinematics model of the robot. To counter the structural feature, the output training samples of networks are improved. The software in which the weights of neural networks are learned by genetic algorithm is designed. And interpolation algorithm is proposed to acquire position and pose of the track’ midpoints. The method mentioned can improve considerably the accuracy of control and realize the pose control rapidly.更多还原