Data-Driven Control for High-Speed Interceptors With Non-Uniformly Sampled Data

This paper investigates the issues of command tracking and attitude stabilization for high-speed interceptors, where inaccurate dynamics information and insufficient trajectory data are considered. With the aid of a classical hierarchical design, the control system of high-speed interceptor is reconstructed into two straightforward feedback loops and an attitude subsystem, the later being characterized by an uncertain time-varying model. For attitude stabilization, an off-policy reinforcement learning (RL)-based method is proposed to estimate the model with only a small amount of non-uniform samples, and a policy iteration (PI)-based approach is utilized to solve the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$ H_{\infty }$</tex-math></inline-formula> controller. By conducting extensive offline computations and employing supervised learning, the mapping relationship between the flight states and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$H_{\infty }$</tex-math></inline-formula> control policies are established, thereby enabling rapid and efficient execution of onboard control decision making. The simulation results validate the accuracy of the dynamics information estimation, demonstrating that the control decision-making mechanism effectively guarantees the attitude stability of the high-speed interceptor while precisely tracking the guidance commands.