Neural network and fuzzy control in the welding robots using CCD camera and touch sensor

It is important to trace the orbit with a corner using a welding robot, in order to obtain a good quality of the welding result. In this paper, a system is proposed for tracking the orbit by using a CCD camera and a touch sensor. When only the fuzzy controller is used, the delay of tracking may occur at the corner. In order to improve the tracking performance, the authors use a feedforward neural network to recognize the corner. According to the recognition results, the control gain of the controller is then adjusted. The validity of the fuzzy controller and the neural network is verified by performing the tracking experiments.