One-Step-Ahead Controller Design using Neural Networks

A one-step-ahead adaptive controller for nonlinear systems using the back-propagation network is presented. At each point of time, the nonlinear relationship between the output and the latest control input of the system is replaced by its first-order approximation. In other words, the nonlinearity is considered to be linear time-varying at each sampling interval, and consequently, one-step-ahead adaptive control can be applied to the nonlinear system considered.