Design of Neural Network Variable Structure Reentry Control System for Reusable Launch Vehicle

A flight control system is designed for a reusable launch vehicle with aerodynamic control surfaces and reaction control system based on a variable-structure control and neural network theory. The control problems of coupling among the channels and the uncertainty of model parameters are solved by using the method. High precise and robust tracking of required attitude angles can be achieved in complicated air space. A mathematical model of reusable launch vehicle is pre-sented first, and then a controller of flight system is presented. Base on the mathematical model, the controller is divided into two parts: variable-structure controller and neural network module which is used to modify the parameters of con-troller. This control system decouples the lateral/directional tunnels well with a neural network sliding mode controller and provides a robust and de-coupled tracking for mission angle profiles. After this a control allocation algorithm is employed to allocate the torque moments to aerodynamic control surfaces and thrusters. The final simulation shows that the control system has a good accurate, robust and de-coupled tracking performance. The stable state error is less than 1°, and the overshoot is less than 5%.