Visually Guided Manipulator Based on Artificial Neural Networks

In this paper, a method is proposed to control the movements of a manipulator using image-based visual servoing, which integrates computational intelligence for object tracking. The objective is to visually servo the 6 DOF PUMA 560 using a charge-coupled device camera attached to the end-effector of the manipulator whereby image is acquired (which is called eye-in-hand camera configuration). The image features are then extracted to form the image Jacobian matrix. The proposed method is based on an artificial neural network (ANN), which is capable of approximating complex and nonlinear image motions to the manipulator motions in order to control the manipulator system. For the visual servoing problem, the gradient descent (GD) and Levenberg–Marquardt (LM) algorithms have been implemented and compared, concluding that the LM algorithm has got better performance than the GD algorithm in visually controlling the manipulator based on ANN. The LM algorithm is integrated into the ANN-based visual control for tracking the periodic moving object in the presence of noise. The simulations show that the controller is noise tolerant too.