| Summary: | 碩士 === 國立高雄大學 === 電機工程學系碩士班 === 104 === Robotic arms are important elements of the automotive industry. The precision of a control system that guides the robotic arm to move to a target location accurately is the key to success of industrial automation. Inverse kinematics refers to the use of the kinematics equations of a robotic arm is a typical method for deciding the parameters associated with the joints that activate the robotic arm to reach to a target location. However, errors from measure inaccuracy, burden, or friction may inevitably exist in the kinematics equations and diminish the precision of the control system. Based on the inverse kinematics, this study develops a practical method for error correction and employs the techniques of artificial neural network (ANN) for training regression models of robotic arms for target localization. The output parameters associated with the joints that produce a precise target location are trained by a supervised back-propagation ANN. Because errors are fixed before being trained by ANN and are considered in the learning algorithm, the control model can guide the robotic arm to reach target location accurately. The proposed method is implemented and validated on a three-axis robotic arm. The experimental results show that our proposed method outperforms typical methods using inverse kinematics.
|
|---|
