Stochastic neural direct adaptive control based on minimum variance optimization

Based on the state space control theory and a neural network architecture, the authors present a stochastic neural direct adaptive control algorithm (SNDAC) for partially known state space nonlinear time varying plants. A neural network is used to generate the control signal, which minimizes a quadratic one-step-ahead prediction performance index based on the minimum variance optimization approach. The SNDAC can be used for both deterministic and stochastic control problems and is computationally efficient and effective. >