Research On Vision-based Line-grasping Control Method For Power Line Deicing Robot

With the development of CCD imaging technology and computer control technology, visual control is largely used in our life and production. The intelligent and autonomous capability of the deicing robot will be increased greatly with the introduction of visual control as for the special application of power line deicing robot. Therefore, the line-gasping control of the deicing robot is deeply researched in this paper.Firstly, this paper reviews the development and current situation of robot visual control. And then introduces the basic concept, classification methods and the typical structures of this area. The deicing robot kinematic and hand-eye model are discussed in the paper. Further more, this paper also proposes the line-gasping control algorithm of two kinds of visual control structure which are position-based visual control and image-based visual servo control respectively.In the research of the basic theory on robot visual control, the differences and advantages of position-based and image-based visual servo control also include the 2.5D visual servo control are highly discussed.The D-H method is used to build the deicing robot kinematics model. And also the hand-eye model is set up to the robot. Then the theory of line-gasping control is introduced in this paper.In the aspect of position-based visual control, a line-gasping control algorithm based on position desired mode visual control was designed. The method puts forward a monocular visual stereotactic algorithm which is based on the cylinder geometry characteristic of transmission line and the CCD image model. In order to avoid complicated computation of the robot inverse kinematics, a line-gasping control strategy is mentioned. The main idea of the strategy is to determine line-gasping point and the desired joint angles of deicing robot when the deicing robnot tries to reach this point through finding out the intersection point in space between the working surface of the end-effctor of deicing robot and axis of transmission lineFinally, the appropriate image features are selected in designing a mixed controller which includes the FCMAC visual controller and the regular PID controller. In this mixed controller, the neural network is trained to the aim of minimal outputs of regular controller. The on-line learning algorithm, which doesn't need to use model information of deicing robot and feature extraction, can help the on-line learning and optimization more quickly and effectively. Thus, the adaptbility of the visual servo system in the uncertainty environment and tasks will be highly improved.Some experiments are done due to the two control methods according to the paper, and the results show that the two control methods are of good control properties.