Unsteady Aerodynamic Modeling and Longitudinal Adaptive Tracking Control for Hypersonic Vehicles with Sweep Variation

Abstract

This paper investigates the longitudinal control problem of a sweep-varying hypersonic vehicle in the presence of unsteady aerodynamic effects, aerodynamic uncertainty, and actuator constraints. In contrast to existing adaptive or reinforcement-learning-based control methods that mainly treat sweep motion as a scheduling variable, the present study develops a unified control framework in which the sweep angle is exploited as an active control input and its nonstationary aerodynamic influence is explicitly incorporated into both aerodynamic modeling and controller design. First, a control-oriented unsteady aerodynamic model is established using the stochastic hierarchical Kriging (SHK) method, and its fidelity is quantitatively evaluated against high-fidelity aerodynamic data through drag-, lift-, and pitching-moment prediction errors. A constrained sweep-angle command and elevator command are then designed to satisfy actuator magnitude and rate limits while preserving longitudinal control effectiveness. To further improve transient performance under uncertain unsteady aerodynamics, an actor–critic reinforcement-learning module is embedded as a lightweight supervisory tuning layer on top of the adaptive controller, so that online policy improvement is achieved without repeated trial-and-error exploration. An observer-assisted implementation is also introduced to relax the full-state measurement requirement. Lyapunov analysis proves the boundedness of the closed-loop signals and establishes the stability of the observer–controller–learning interconnection. Numerical results from nominal ascent/descent tracking, disturbance rejection, RL-ablation comparison, and Monte Carlo shooting simulations show that the proposed method improves aerodynamic prediction fidelity, reduces tracking error and settling time, and enhances robustness relative to baseline adaptive and PID schemes.