Intelligent control of air-breathing hypersonic vehicles subject to path and angle-of-attack constraints

This paper investigates the path and angle-of-attack constrained longitudinal control problem for air-breathing hypersonic vehicles with the help of a two-step intelligent design. The first step generates the path constrained optimal trajectory via a deep neural network, while the second step develops a reinforcement learning-based controller. In the velocity loop, an adaptive tracking controller is designed with specific consideration on the constrained fuel-to-air equivalency ratio of the scramjet. Meanwhile, in the altitude loop, the baseline backstepping design is enhanced by an intelligent parameter adjustment to improve tracking performances as well as to constrain the angle-of-attack, without involving any complex mechanisms like the widely-used prescribed performance functions and barrier Lyapunov functions. Numerical simulations show both effectiveness and superiority of the proposed intelligent control.