NONLINEAR STABILIZATION BY RECEDING-HORIZON NEURAL REGULATORS

A receding-horizon (RH) optimal control scheme for a discrete-time non linear dynamic system is presented. A non-quadratic cost function is c onsidered and constraints are imposed on both the state and control ve ctors. A stabilizing regulator is derived by adding a proper terminal penalty function to the process cost. The control vector is generated by means of a feedback control law computed off-line instead of comput ing it on-line, as is done for existing RH regulators. The off-line co mputation is performed by approximating the RH regulator by a multilay er feedforward neural network. Bounds to this approximation are establ ished. Algorithms are presented to determine some essential parameters for the design of the neural regulator, i.e. the parameters character izing the terminal cost function and the number of neural units in the network implementing the regulator. Simulation results show the effec tiveness of the proposed approach.