얇은막대 배치작업을 위한 N-R 및 EKF방법을 사용한 로봇 비젼제어기법 개발

Abstract

This study uses the vision system model which is proved in previous studies to efficiently transform the three-dimensional position values of the robot to the two-dimensional camera plane values using the software method. The vision system model has six camera parameters. Among six parameters, the first four parameters indicating the camera inner parameters, explain the uncertainty of the camera's orientation and focal length, and the other two parameters indicating the camera outer parameters explain the uncertainty of the relative position between the camera and the robot. In order to increase the precision of the vision system model, the positions of the cameras are selected so that the object attached to the end-effector of the robot is filled in the image plane of each of the three cameras while the robot moves along the arbitrary motion trajectory. In particular, to minimize the amount of data, the test model using an LED-cues representing the entire object shape is manufactured. Thus, using this vision system model, the two robot vision control schemes, based on the N-R and EKF methods, are proposed for object placement task to the fixed target point. The spatial position value of this target is unknown but only the vision data is known. The experiments of slender bar placement with two cues are performed in order to show the effectiveness of the proposed two vision control schemes. From the experimental results of slender bar placement involving positional accuracy and processing time, we would like to identify the advantages and disadvantages of the proposed two robot vision control schemes with different numerical analysis characteristics. Also, for the judgement of the vision system model's fitness in EKF method, the changes of the estimated six camera parameters used in the vision system model are observed, and their characteristics are identified. Additionally, the robot inverse kinematic analysis is carried out for the estimation of robot's joint values to the fixed target with the known spatial position value, in order to set the criterion for the judgement of positional accuracy of the two robot vision control techniques.