Görsel bilgi ile robot manipulatörlerin adaptiv kontrolü

ABSTRACT ADAPTIVE CONTROL OF A ROBOT MANIPULATOR BY VISUAL DATA KAPUCU Sadettin Ph.D. in Mechanical Engineering Supervisor: AssocProf. Dr. Sedat BAYSEÇ December 1994, 142 pages One important research area on robotics is equipping robots with more sophisticated sensory systems such that control computer compiles more extensive data about the environment and the manipulator itself. This extensive data help to monitor the events occurring, to adapt and to respond accordingly. Among the senses known to man, vision is by fer the most important, and helps not only in data compilation but learning as well. Although vision provides an invaluable data, generally its processing takes too much time, not permissible at the shop floor. This thesis aims at putting simpler means of determining the position of an object, defining it not by intricate image identification techniques but by simply specifying it based on 3 non-coinciding spots of distinctive color put on the objects handled, by photogrammetry in stereovision. A set of general purpose control routines are developed to control any size and topology manipulator. The type of joints and link parameters are obtained by incrementing the joint motors and monitoring the corresponding motion of the manipulator tip. Location of the end effector is identified iiiby video cameras searching for a spot of distinctive color at the tip. From this data, equations of forward and inverse kinematics are derived in a systematic and symbolic manner. To identify the dynamics of the manipulator, system identification techniques are used which assume second order linear differential equations for the motion of the joints, coefficients extracted from the response to binary steps, off-line. Lastly, a control program is presented which provides step commands to a manipulator to reach an object whose position is determined visually. The ideas generated are tested on a hydraulic, heavy duty robot manipulator, considering only 3 gross degrees of freedom, providing point positioning in space. This thesis is the major part of a research project entitled Robotlarda Görsel Bilgi Analizi, carried out at Mechanical Engineering Department of Gaziantep University, with the financial support of State Planning Organization of Turkey with project number 91 K 120 830. Keywords: Robot control, stereovision, sistem identification. IV