Adaptive neural network control of flexible manipulators

A back-propagation neural network is applied to the end-position control of a one-link flexible manipulator with payload variations. It is challenging to control complex mechanical systems such as flexible manipulator and to find computationally acceptable methods to compensate for variations in parameters such as payload of the system. Neural networks are capable of handling the computational complexity which sometimes can not be described in detailed mathematical terms. A deadbeat controller tuned by neural network for a single flexible arm robot is presented to demonstrate the capability of neural networks in the field of robotic control. In such an application, the neural network is employed to adjust parameters of the deadbeat controller as well as to identify changes of system dynamics when the flexible arm robot is manipulating objects under varying payload conditions.