Visual Servo Of Manipulator To Keep Image Features In FOV

As the robotic techniques developing, robot becomes more and more important in the normal life and industrial area. To apply the robot in the non-structural environment, it is necessary to acquire the exact position of the objects. Thus it is impossible to control the robot merely by classical trajectory planning method. Computer vision, which can work as the human eye do, can provide the comparatively strong ability of recognizing the surround environment. For this reason, more scholars pay attention to the visual servo control of robotics.In the visual servo control, how to acquire the visual information is significant. Visual servo can be classified by which place the camera fixed in as two different system:Eye-in-Hand system and Eye-to-Hand system. In the Eye-to-Hand system, the position of the camera is invarious, and it can observe the whole working area. But the image information may be non-precise; In the Eye-in-Hand system, the camera is normally fixed in the end-effector of the robot(manipulator). It moves as the end-effector moves.Thus the image information may be more precise. But as the camera moves, the image features may go out of the field of view(FOV). To keep the image feature points in the FOV, this paper does research on the visual servo and manipulator kinematic control. The details is as the following:Firstly, this paper analyzes the basic control method of visual servo, and builds a platform to realize the dual-camera visual servo of two degrees manipulator. In addition, the paper collects some method of keeping image feature points in the FOV. Inspired by them, we would like to regard the problem of keeping image feature points in the FOV as the subtask in the manipulator control.Secondly, the paper illustrates some important method of kinematic control of redundant ma-nipulator, and emphasizes the concept of visual joint. Further more, to revise the shortcoming of General-Weighted Least-Norm method (GWLN) which can lead to frequently steep change, a new weight setting mechanism is provided.Thirdly, based on the concept of visual joint, a virtual image joint method is provided. As the same as real joint, virtual image joint has lower and upper limits. So it is easy to apply the limits and avoidance techniques to keep the image feature points in the FOV.Finally, based on the redundant manipulator with seven degrees, a visual servo system is built up to implement the experiment after operating MATLAB simulations. Then the effectiveness of virtual image joint method is proved.