A neurocontroller for a biped robot

A neural network architecture is presented for the control of a three-link biped walking robot. The proposed neuromorphic controller, based on a hierarchical structure of artificial neural networks, was trained by supervised learning. The training model was derived by applying nonlinear feedback decoupling and an optimal tracking strategy. The neurocontroller utilizes several useful features of neural networks, such as generalization, parameter adaptivity, and robustness. Based on a comparison of performance obtained with an optimal control law and that obtained with the neurocontroller, it was concluded that the neurocontroller provides superior performance in the presence of large disturbances