Abstract: The application of neural network technique in the seam tracking process of arc – welding robot is studied. A neural network controller to modify the reference Cartesian trajectory for the robotic seam tracking is presented through the compensation of the neural network for the Cartesian trajectory. Compared with the traditional computed torque control method, the proposed neural network controller yields salient controlling characteristics and strong robustness. The precision of seam tracking is conspienously promoted.

Key words: Neural network, Robot, Seam tracking