Vision Based Robot Manipulator Control in 3D Space Using Fuzzy System and Neural Modeling

Visual servoing system controls a robot by visual feedback so that robot moves from any arbitrary start position to the target positions. The coordinates of points in three dimensions is needed in 3D space. In this paper, a Kinect camera is used to collect RGB images as well as workspace’s depth matrix. The control law is obtained using Jacobian matrix. Since, the mathematical model of robot and workspace, is unknown, artificial neural networks is applied to approximate inverse of Jacobian matrix by gathering data. The approximated neural models are used in control law directly. For each degree of freedom of the robot manipulator, a two-layer feedforward neural network is considered. The distance between end-effector and target in 3D space, and the shoulder joint coordinates are inputs of each of the networks and outputs are the fraction of the related robot joint changes to the image features changes (the elements of inverse of Jacobian matrix). The proposed method has been implemented on an industrial robot manipulator. The experimental results show that the proposed control system can move the end-effector to different target positions in workspace with good accuracy and fewer steps in comparision with the previous method.