A neural network adaptive inverse controller for hypersonic vehicle

A structure of nonlinear adaptive inverse controller based on Radial Basis Function (RBF) network is proposed to control a hypersonic vehicle with highly nonlinear and strong coupled states. Different from conventional adaptive inverse controllers, the proposed one is constructed by one main controller called Trimmed Adaptive Inverse Controller and two compensators called Attitude Angle Compensator and State Compensator respectively. The Extended Minimum Resource Allocating Network (EMRAN) algorithm is adopted to train the RBF network on-line and off-line. The simulation results show that proposed adaptive inverse controller could lead the hypersonic vehicle to trace the expect input rapidly and steadily, and could adapt the different flight condition by online learning.