Rigid model-based neural network control of flexible-link manipulators

Applications of neural networks to the identification and control of flexible manipulators are considered. Usually, neural networks need a large number of neurons and a great amount of computation for learning, and the error is not easy to reduce. This study tries to combine the a priori knowledge of the corresponding rigid manipulator's model with two multilayered neural networks for the identification and control of a flexible-link manipulator. The suggested approach can use fewer neurons and needs shorter learning time for reducing the error. A planar (in the vertical plane) two-link flexible arm with the first link rigid and the second link flexible is tested via simulation. The mathematical model of the flexible arm for simulation is derived by the finite element method using Lagrange's equation.