Research On Robot Uncalibrated Visual Servoing System Based On Multi-Intelligent Control Methods

The robot visual servoing system is an import branch of the modern intelligent robot research and has broad application prospect, because it simulates the biologic untouched vision information sensing and feedback mechanism and has the adaptability to the environment and object. On the other hand, the robot visual servoing system is also a difficult and challenging research project, because it is involved in a lot of subjects such as robotics, dynamics, image processing, control theory, computer science and has the characteristics of nonlinearity, deep coupling and difficulty for modeling. Above mentioned factors make the traditional methods based on the exact models and eye-hand calibration hard to achieve satisfactory control effects or possess universality in application, as a result, the robot uncalibrated visual servoing ststem is becoming one of the main research directions. Considering these reasons, several novel methods based on intelligent control are studied and presented to realize the robot visual servoing control without calibration in this paper.Firstly, the main image processing especially the image segmentation, the features choosing, extracting and matching technology in visual servoing is analyzed. Considering the importance of image segmentation in the algorithms of this paper, a new segmentation method based on two-dimensional entropy and Genetic Algorithm(GA) is presented in this paper, which improved the segmentation speed greatly by making use of the features of intelligent searching, parallel processing and adaptive learning supplied by GA while maintaining the advantages of the two-dimensional entropy method such as the excellent stability and high accuracy.Secondly, a novel control method based on artificial immune evolutionary algorithm (AIEA) and image error is proposed in this paper. Because of using all the image information and avoiding the image features choosing and extracting, the better universality and stability can be realized based on the technology in this paper. The experiment results show that the method is valid and practicable.Thirdly, aiming at the solution of the difficult and complicated marking, extracting and matching for the image geometrical features, A visual servoing of 4DOF is proposed based on the image moments and back progagation neural network (BPNN). Because of using the global image moments descriptors and the direct mapping between the image moments variation and the robot pose displacement, this method has a lot of advantages such as for simpler image feature error computing, no eye-hand model and the exteral and interal parameters calibration of the camera needed. The experiment results verify that the method is easy to implement and can acquire satisfactory servoing performance.Finally, for the problem of visual tracking of a moving object, a visual servoing method based on the parameters adaptive fuzzy congtrol (PAFC) and 2-1/2D architecture is proposed in this paper. To realize the task, the whole system is decomposed into two subsystems, one of which is used to implement the translation tracking by PAFC and image-based visual servoing (IBVS, or 2D) structure, the other perform the function of rotation tracking by PAFC and position-based visual servoing (PBVS, or 3D) structure, and the advantages of both the IBVS and PBVS are utilized by the new method of this paper as a result. The conflict between the dynamic rapidity and static stability can be solved successfully by the adopting of the parameters adaptive fuzzy control.In the meanwhile, simulation and experiments for every kind of control method are carried out to verify the efficiency and validity.