An experiment in approximating an end effector positional error of a 6 D.O.F manipulator using neural network

In this paper, a BP network is developed to approximate an error vector in the joint displacement of a 6 D.O.F. Stanford manipulator. This error results from several causes such as physical damage to the robot structure or inherited inaccuracies in the robot design. Results show that using neural networks is a robust and efficient solution.