Design of artificial neural network predictor for trajectory planning of an experimental 6 DOF robot manipulator

Abstract

Nowadays, the use of robots is continuously increasing in industry. Especially, robotic Gas metal arc (GMA) welding is widespread used as manufacturing process. Because of this increase in the use of robots in the industry, there is a need to study on a number of improvements. This paper presents an experimental research on the robot manipulator, using image processing to detect location of welding seam for the planning optimal trajectory. This new study provides the weld seam trajectory to be created without being affected by manufacture faults. Firstly, communication interface between the robot and the computer is developed by using previous related software library. Then, the weld seam trajectory are automatically generated using image processing via Matlab and reference points are determined on the trajectory for tracking of the manipulator. The values of this points are sent to the robots for calculation of the joint angles by the software on the robot side. Furthermore, the related parameters are tested with neural network predictor to predict optimal trajectory on resulting image using image processing. The results show that this approach improved that neural network predictor can increase trajectory accuracy for quality welding process.