Hypersonic aerocraft neural network control method based on online data learning

The present invention discloses a hypersonic aerocraft neural network control method based on online data learning. The objective of the invention is to solve the technical problem that a current hypersonic aerocraft control method is bad in stability. The technical scheme comprises: the longitudinal channel model of a hypersonic aerocraft is decomposed into a speed subsystem and a height subsystem, a neural network performs approximation of the nondeterminacy existed in the hypersonic aerocraft system, and online data is employed to construct prediction model error of a neural network to perform neural network weight updating to improve the system learning capacity of a neural network closed-loop control process. Aiming at the two subsystems, a controller based on neural network learning is given out to realize nondeterminacy hypersonic aerocraft dynamics control. Good or bad neural network approximation nondeterminacy learning is described to the composite update of the model errors introducing the neural n