Generalized stochastic state space neural adaptive control

A generalized stochastic neural adaptive control algorithm is presented, where the system identification is based on the state space innovations model and a neural network architecture. This identification algorithm is derived from three different optimization approaches, i.e., the gradient, Newton, and minimum variance. The control law, also based on neural networks structure, is derived from a quadratic (one-step-ahead prediction) performance index, which in combination with neural identification constitutes a unique neural adaptive control algorithm with excellent performance.<
>