In many situations controlling some robot manipulators or other plants, we have to struggle with the 'inverse problem'. This paper propose a new framework of neural network system witch can obtain both the direct model and the inverse model of the plant simultaneously, and shows that using this framework, both of the efficiency and accuracy of inverse models, are improved. We are now investigating the possibility of the application of this framework for solving inverse kinematics of robot manipulators.
Bi-directional system modelling technology and its application for controlling robot manipulators
