Real Time Trajectory Generation Method and Control of a Sealing Robot : Prediction of Desired Position of a Visual Sensor Using Off-Line Learning

Abstract

In the previous papers, a sealing nozzle and a visual sensor were mounted on the end-effector of a manipulator, but control of nozzle position was only considered. As the results, the visual sensor would possibly lost the seam target at curved part of a sealing line during tracking. In order to improve tracking property of the sealing robot, both a visual sensor and a nozzle of a sealing robot must be real time controlled. In this paper, a fuzzy model based method for generating desired trajectory of the visual sensor is proposed. The fuzzy rules are obtained by applying neural network learning. Generated desired positions of nozzle are equivalently converted to rotation angles considered in the visual sensor coordinate. Then, a visual feedback algorithm for simultaneously controlling the nozzle and the visual sensor is also developed. Using the proposed method, positions of the sealing nozzle and the visual sensor can be controlled at every sampling instant in the robot control system. Therefore, both the nozzle and the visual sensor can move along a sealing line without losing the target during tracking. Tracking experiments for sealing objects with different type of sealing lines are carried out, and the effectiveness of proposed method has been demonstrated.