Optimising Visual Servoing to Avoid Robot Arm singularity

Abstract

The thesis presents the position based visual servoing control scheme which use a visual feature in the control loop as a feedback signal to control a robotic arm to reach a desired position. The con guration used for combining the visual sensor and the robot is called eye-on-hand con guration where the visual sensor is attached to the robot end e ector. A CRS 5 DOF was chosen to perform the visual servoing task. A complete and clear analysis for the robot kinematics focusing on it's singularities was studied and pointed out clearly. The e ect of reaching singularity on the robot components was shown in a real time experiment. It is important to avoid such singularity e ects during visual servoing tasks. That is why in this work we used genetic algorithm to optimize the location of the end e ector when it is around such singular positions during task execution.