Intelligent back-stepping longitudinal control of hypersonic vehicles

Aiming at the longitudinal flight of air-breathing hypersonic vehicle (AHV) with multi-constraints, the traditional back-stepping design with advanced intelligent techniques are combined to develop the flight control system. To consider the complex constraints and performance indexes, a deep neural network is constructed and trained based on numerous trajectories optimized by the Gauss pseudo-spectral method, which can generate the nominal trajectory in a rapid manner. Moreover, a simple nominal back-stepping controller is designed for the longitudinal dynamics of AHVs. Key parameters of this back-stepping controller are adjusted online with the help of the deep reinforcement learning, so as to further improve the control performance. The proposed intelligent back-stepping control strategy well adapts complex constraints, while having a relatively simple structure. Numerical simulations preliminarily show the effectiveness of our proposed method.