Abstract

Nowadays, industrial robots are widely used in many fields. With the wide range of applications, the accuracy of robots has become an issue of concern. In some path following tasks, the accuracy of existing robots does not yet meet the industry standard. In this research, the accuracy of the robot is enhanced by using position-based visual servoing. The purpose of this research is to propose a novel dynamic path tracking (DPT) method to solve the aforementioned accuracy problem. This method uses the C-Track from Creaform to measure the end-effector of the robot M-20iA from Fanuc and to use the visual information to guide the robot to follow the desired path. First, a stereo binocular camera C-Track provides the real-time pose information of the end-effector. To remove the noise transmitted from the camera, the research uses a robust Kalman filter (RKF) to improve the performance of the standard Kalman filter under disturbance by correcting the state variable error covariance (P). Then, the Fanuc robot M-20iA uses position-based visual servoing (PBVS) strategy to correct the position and orientation of the end-effector, which is implemented through dynamic path modification (DPM) function, in conjunction with real-time data acquired by C-Track. Next, adaptive neuro-PID (ANPID) control is developed as the PBVS scheme for DPM correction. Such control strategy has a strong adaptive and self-learning capability, which enables online tuning of the PID controller parameters, resulting in better performance in robot control. Finally, extensive experiment tests have been carried out and the results show that the the accuracy of path following reaches ±0.08mm and ±0.04deg, compared with the accuracy ±0.2mm and ±0.1deg achieved by conventional PID controller.