Research On Robot Visual Servoing Control Based On Position

Using sensor information to control robots has become a very popular field of research, since it promises to lead to the design of autonomous robots, In contrast to their preprogrammed industrial counterparts, autonomous robots must be able to deal with unexpected events such as obstacles or misplaced objects. Nowadays, vision is by far the most commonly used sensor because CCD cameras are cheap and easy to use. Additionally, vision is an important human sense and information received by the robot's vision system is easier for the human operator to understand, hi the field of visually controlled robot manipulators, research has mainly focused on the control part, circumventing the problems of extracting and interpreting image features by using artificial features. This resulted in a large number of impressive servoing methods that cope very well with a specific problem.Having detected the target object in the video image, the question arises which visual information is to be used to control the robot. The position-based approach first estimates the position of the target object relative to the camera in Cartesian coordinates, based on a geometrical model of the robot, its reachable work space (task space) and the target object. The error signal for the robot controller is therefore defined in Cartesian coordinates. As most robots provide a Cartesian interface and because operating in 3D space can be understood intuitively by the system designer. However, an exact determination of the pose of object to grasp relative to the manipulator requires an accurate hand-eye calibration and precise pose estimation.In image-based, also called feature-based systems, the error signal is defined in terms of image features and is directly measured in the image coordinate system. Therefore, computational costs are significantly reduced and the system becomes less sensitive toABSTRACTerrors in camera calibration and system modeling. However the computation of robot motion on the basis of image features takes place in a less intuitive projection of the task space, depending on the chosen image features and the method determining the distance between the features. As this process is non-linear and its parameters highly correlated, it presents a significant challenge to control design and has proven to be difficult to analyze theoretically.The thesis is mainly concerned with fundamental configuration and implementation methods of the position-based visual servoing robot system. Improvement is made on the target recognition in image by pattern match technique based on Step GA. Target recognition is more accurate and faster by combining the local GA and global GA, thus greatly improving the performance of the visual servoing system.Considering that the noise caused by image acquisition and transmission may possibly degrade the overall system performance, we present a new fuzzy-neural network image filter. The thesis discusses intelligent control over robot manipulators and formulates a hybrid controller as well, which combines fuzzy-neural network with CMAC controller and slide mode controller.Here lists the innovations and achievements in the research.1. Target recognition in image by pattern match technique based on Step GA is presented. In order to satisfy dynamic characteristics of the robot system, the individual that has the maximum fitness value is assigned to be input of the robot visual controller by evaluating inverse kinematics after each generation GA evolution. So the path planning based on GA process is considered to be in real time mode, which satisfies the requirements of a real time dynamic system. The research on real-time recognition of road traffic sign based on genetic algorithm and Simple Vector Filter (SVF) has also been completed.2. The method combining Local GA with Global GA optimizes the performance of visual servoing system. First, target recognition preprocess is carried out by global GA. After the fitness value reaches a certain threshold value, the Global GA will...