Robust online MPC tracking of tychastic hypersonic trajectories via decoupled EKF–DSF estimation

Hypersonic reentry vehicles face strict aerothermal constraints and severe environmental uncertainties. This paper addresses the problem of robust online tracking under double uncertainty: significant initial state injection errors combined with extreme atmospheric density mismatches. It is first demonstrated mathematically that traditional coupled estimation schemes are ill-conditioned for this task due to instantaneous unobservability and extrapolation error amplification. To overcome these theoretical limitations, a decoupled architecture is proposed, comprising a kinematic extended Kalman filter (EKF) and a dedicated density scale factor (DSF) estimator. This multi-rate estimation scheme feeds a reduced-order model predictive controller (MPC) that tracks an offline-generated robust reference trajectory. By leveraging the inherent probabilistic safety margins of the reference trajectory, verified via Cantelli’s inequality, the proposed MPC maintains thermal safety without the computational risk of active online constraint management. Extensive Monte Carlo simulations verify the effectiveness of this architecture. While a baseline controller relying on nominal atmospheric models produces large terminal deviations with mean altitude errors exceeding 4 km, the proposed architecture provides improved tracking performance with a mean error of approximately −63.8 m. Crucially, no statistically significant difference was detected between this performance and that of a theoretical system with perfect atmospheric knowledge (mean error (Formula presented) m). These results indicate that the decoupled framework isolates kinematic noise from parameter bias, providing a viable path for safe, autonomous hypersonic flight.