Robot arm guidance using visual servoing

• Main Authors: Dealino, Francis Michael, Dee, Brian John Anthony, Payuyo, Philip Andrew, Ponce, Jackie, Redor, Karla Vanessa
• Format: text
• Language: English
• Published: Animo Repository 2004
• Subjects: Robots > Control systems; Computer vision; Servomechanisms
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/14532
• Institution: De La Salle University

This study generally demonstrates the concept of visual servoing. Visual servoing is very helpful in improving the potentials of robots found in many industries. Instead of adjusting the work environment to suit the robot, such concept enables the machine to manipulate its environment and not the other way around. In order to demonstrate the concept, a specific type of visual servoing architecture must be chosen: in this study, dynamic position-based look-and-move. This type of visual servoing architecture uses features on the captured image and relates it to the known model of the object to determine its relationship to the camera. To show this, we used a checkerboard as object concerned in the study. The user specifies a target spot relative to the object frame. Points defined in the checkerboard serve as the feature that necessitates the information needed to establish the relationship between the camera and the object. By the use of principles in kinematics the pose of the end effector will be determined to reach the target. Along the way, continuous visual feedback is present to establish its current pose to the desired pose. We use the UMI RTX owned by the department as the robotic platform for the study. The robot's electronics and mechanical components are retained except for the gripper. The RTX library written in C ++ provides the main channel of communication between the robot and the PC. With regard to the image processing, the functions from the OpenCV library are used. The kinematics component of the software used in the study is also developed using the functions from the said library. An algorithm called COPOSIT is used to acquire the relationship between the camera, the object and the robot's end effector. To explicitly justify the concept, experiments are done to inspect the overall performance of the whole application. Some experiments are responsible for building the main concept of the study while the others are performed to quantify the possible sources of inaccuracy that affect the performance of the system. The results are presented and evaluated accordingly. Finally, based on the data gathered and interpreted, conclusions are drawn to check whether the objectives are achieved or not. Recommendations are then listed to improve the aspects of this study that need further research.