الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 16 |  |  |
| | | from | 16 | | |
Abstract
Visual sense is one of the most human senses that provide sufficient information and non contact measurements from unknown environment. Without visual information, many arm robot tasks can be only performed in a well known environment where every object is known and can be found in a well known pose. Any error in the pose of the object or about the robot pose will lead to task failure. Potential sources (such as gear backlashes, bending of the links, joints slippage, and poor fixture) would lead to the errors in the robot end-effector pose.
Position based visual servoing control system consists of many subsystems such as vision sub system , robotics sub system and computer sub system which are combined together in one control scheme. Position based visual servoing control system is used to solve the above problems by using the visual information in a feedback loop to control the robot end-effector pose relative to the pose of the object being manipulated. The main core of position based visual servoing (PBVS) is the estimation of the pose of the object frame with respect to the camera frame. Many algorithms have been developed to solve the problem of pose estimation.
To implement this work, the theoretical study of robot kinematics, vision systems, pose estimation methods and visual servoing control schemes is very important. So, it was the first step of this work.
The second step of the work is the simulation. It is a vital step because it can help in the prediction of the overall system behavior. And also can help in saving our platform from any dangerous motion through the experiments. CRS robot simulator and real camera simulator are developed and combined through the control scheme to test and evaluate the pose estimation methods and the PBVS system.
To develop simulator that accurately mimics the real system, the actual system parameters must be known and used through the simulation. So camera calibration is performed to measure the camera intrinsic parameters also the camera robot calibration is performed to calculate the actual transformation of the camera frame with respect to the end effector frame.
Working on the real environment is the last step. In the real environment, the open architecture controller of the real robot has been studied and modified to be suitable for combining with the PBVS control scheme.
In the real image, certain features must be extracted and arranged in the same order of the previous image features to be fit to be used as input of pose estimation software. So, feature extraction and feature matching functions must be used to prepare the real image.
Kalman filter and classic Posit are two techniques that are used for pose estimation process. The two methods are implemented and tested through the simulation. The algorithm with the best results is used in the PBVS control scheme.