A neural network based receding horizon optimal (RHO) controller

A neural network based RHO controller is developed for jet aircraft engines. It takes advantage of the learning ability of the neural network to obtain the mapping function between system input and output, and does not predicate upon a priori knowledge of the system model. The controller was tested using OREOX, a jet engine simulator provided by Pratt and Whitney. The controller recovers from system changes in seconds. Due to the smoothing and stability measures undertaken, the control trajectories are smooth and stable even when the target thrust is changed abruptly.