激光加工机器人标定的神经网络法

在分析原有机器人误差标定方法的基础上,给出了两种误差标定的神经网络法,并对激光加工机器人进行了精确标定。第一种方法通过前馈神经网络与机器人理想运动学模型结合,可把机器人位姿误差降低到初始值的1/5左右;第二种方法把前馈神经网络与标定好的实际运动学模型结合,进行神经网络-参数法混合位姿标定,并与参数法标定结果进行了对比。仿真结果表明后一种方法可以进一步提高机器人的位姿精度,且可以对参数模型中没有涉及到的因素进行补偿,方法简单且标定效果更好。

Calibration of Laser Processing Robot by Neural Network

Based on the analysis of the traditional robot calibration methods, two neural network calibration methods used to calibrate the laser-processing robot were proposed. The first one combines a feed-forward neural network with the nominal kinematic model of the robot to be calibrated; which can reduce the pose error to 1/5 of initial value; and the second variant combines a feed-forward neural network with an already calibrated parametrical kinematic model of the robot. The latter reaches a higher performance. The presence of the neural network permits the compensation of several effects, even those not considered by the parametrical model. Calibration results were compared with those obtained by traditional parametric methodologies. Simulation results show that presented methods significantly improve and achieve better calibration effect.