Artificial Neural Networks Based Manipulator Visual Servoing

چکیده مقاله

In this paper, an artificial neural network (ANN) based controller is proposed in order to visually control the 6 DOF PUMA 560 manipulator. A CCD camera is attached to the end-effector capturing images from robot task space. Image features are then extracted to form the image Jacobian matrix. The relations between image features movements and manipulator movements are highly nonlinear and complex. ANNs are a good choice to learn the aforementioned relations due to their generalization and learning nonlinear relations ability. They are also noise tolerant so in the presence of noise, they can control the system pretty well. The proposed ANN based visual controller using a three layer feed-forward perceptron with supervised learning Levenberg-Marquardt (LM) algorithm is trained. Different experiments with different neurons in hidden layer are also carried out to obtain the best performance. Eventually, the ANN based controller is applied to image based visual servoing (IBVS) system for tracking sinusoidal moving object in the presence of noise. The simulations show outstanding results in tracking sinusoidal moving object in the presence of the noise.

دریافت لینک دانلود مقاله