Home Curriculum Vitae Publications Dr. Theodore Pachidis's Home Page Ph.D. Abstract
|
Industrial robots play a key roll to production systems. However the most of them, which are systems of closed architecture, do not use sensors and mainly vision systems. Moreover, communication with these systems is restricted. Thus, they are less flexible, reliable and less adapted to the requirements of an industrial environment. In this dissertation, systematic efforts are made to exploit the minimum communication facilities provided by a robotic system and at the same time, robotic control, in real time, by means of personal computers (PCs) and Windows of Microsoft as operating system (OS) to be achieved. PCs are the most common computer platform while Windows are the most common OS. The objective of using the above hardware and software is to create robotic systems user friendly that will be possible to serve in smaller production units. A novel architecture, called HumanPT, is presented. By means of this architecture, PC-based robot control is achieved by using sensors and particularly vision systems. In the frame of this dissertation, a complete robotic application (HumanPT) in visual C++ based on the above architecture was developed. It is implemented on a Unimation PUMA 761 robotic manipulator. The objective was to create an arc-welding robotic system mainly by means of the Pseudo-Stereovision System (PSVS) proposed. The communication of a PC with the robot is achieved through PUMA ALTER serial port for first time at 38400 bps. PSVS, which is composed of a camera and four mirrors, captures in one shot a complex image as result of the superposition of the left and right view of PSVS. Construction details and the related equations are provided, refraction phenomena due to the beam-splitter used are studied and equations giving coordinates of a 3D point, where refraction phenomena are taken into consideration, are recalculated. Separation methods of complex images are also studied (for color and gray-scale complex images). To reconstruct each pair of images generated by the separation, the meander method of reconstruction is proposed. A novel corner detection method, implemented to 2D objects used by a camera calibration method, is presented. The algorithm is based on seeds, while corner detection is achieved by means of X-shape 11X11 templates. The related theory and the environment, part of HumanPT application, permitting camera calibration and hand-eye calibration, in a short period of time, by means of known calibration methods, are presented. Two original path generation methods are also presented. The first of them is based on complex images captured from static scenes. By means of a number of original algorithms, the generation of simple or of more complicated paths is achieved. For path generation, objects edges (lines or curves) and lines designed on complex images by using the proper tools provided are utilized. Some of the original algorithms proposed are, a correspondence algorithm for complex images, a successive color edge points algorithm and a thinning algorithm for color edges. The second method is based on pseudo stereovision visual servo-control (PS-VVSC) and an original target-object (TOB) constructed for the requirements of this method. The method permits path generation in the whole robot environment. Finally, a novel control unit generation software mechanism from structure elements, and a methodology for the final desired path generation is presented. A real time control method, in which the desired path is optimized by means of PSVS, is also provided. Control schemes based on the corresponding control units are given and their operation is explained. The whole system developed is extendible, easy adapted to other robotic systems and permits in an easy way to generate a desired path and then to drive the robot end-effector along this path for many different tasks or procedures. It can be also used in mobile robots.
|
| Home Curriculum Vitae Publications Last Updated
21/3/2021
|
