Multi-model optimal controller design for Near Space High-speed Vehicle

Attitude control is one of the key technologies for Near Space High-speed Vehicle (NSHV) which has great strategic significance. The NSHV has wide flight envelope with fast time-varying, strongly nonlinear and seriously coupling dynamic characters challenging the attitude control in task-adaptability greatly. Based on singular perturbation theory, fast and slow loop controllers are designed by backstepping technique to accomplish attitude tracking. A multi-model optimal control method is presented as the attitude control law. Subsequently, a radial base function neural network (RBFNN) is applied to approximate the multi-model optimal state feedback gain matrix space. So the multi-model optimal controller (MOC) achieves real-time accurate vehicle attitude control. The traditional gain scheduling (GS) method depends on the model and its procedures are cumbersome. Nevertheless, the MOC is simple and easily realized, overcoming the disadvantages of GS. The simulation studies demonstrate that the proposed MOC meets the performance requirements with acceptable control input. And the design of the MOC possesses effectiveness and engineering value.