论文部分内容阅读
Since the appearance of the neural networks, the industrial environment has jumped a big step toward the future, most of the processes that acquired a human supervising in the past , has now their own learning algorithms, one of the most influenced fields by this is the Robots, I the near past most neural algorithms developed were for robots control, the actual objective is to create a robot that could have the agility of human, and this could be accomplished in the near future due to the fast developing of microprocessors, and the appearance of supper computers.One of the most challenging application in this field is the visual control, knowing that it’s done in real time, make it even more difficult than any of the other method, most of these algorithms use two visual sensors to provide the system with a 3D scene, that contains the necessary data to such as the position of the objects and their shape, which is identical to the work of he human eyes.In this thesis we will discuss and realize one method that uses the neural network to generate the tasks that the robot should do. In the first chapter we present a general view on the application of neural networks for controlling robots com, and as the control is visual we need to develop an image processing algorithm to receive the necessary data from the sensors, and this is the subject of the second chapter. The chapter 3 discuss the neural networks, their principles and some of their most important training algorithms. In chapter 4 we discuss the problem of coupling the visual system with the robot, and implement the command system that we propose, the next part was truly the most interesting, it was some kind of challenge, the chapter 5 discuss the realization of two 4 degrees of freedom robot arms, that could be controlled from PC which make it easier to implement the controller we want, and the last part is the discuss the results of our design, and take a final conclusion, on the system, while the appendices A contain several issues in programming with the visual C++, B present the hardware issues, such as the sensors calibration...etc, C encloses the list of the tables and the figures, while D contains the bibliography.