Entry Guidance for Hypersonic Glide Vehicles via Two-Phase hp-Adaptive Sequential Convex Programming

• Published in: Aerospace Vol. 12; no. 6; p. 539
• Main Authors: Liu, Xu, Li, Xiang, Zhang, Houjun, Huang, Hao, Wu, Yonghui
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.06.2025
• Subjects: Accuracy; Actuators; adaptive mesh refinement; Algorithms; Approximation; Atmospheric entry; Collocation methods; Convex analysis; Convexity; Design and construction; Efficiency; entry guidance; Error reduction; Geometric constraints; Gliding and soaring; Grid refinement (mathematics); hypersonic glide vehicle; Hypersonic planes; Linear quadratic regulator; Objective function; Optimization; Real time; real-time trajectory optimization; sequential convex programming; Trajectory planning; Vehicles; China
• ISSN: 2226-4310; 2226-4310
• DOI: 10.3390/aerospace12060539

This paper addresses the real-time trajectory generation problem for hypersonic glide vehicles (HGVs) during atmospheric entry, subject to complex constraints including aerothermal limits, actuator bounds, and no-fly zones (NFZs). To achieve efficient and reliable trajectory planning, a two-phase hp-adaptive sequential convex programming (SCP) framework is proposed. NFZ avoidance is reformulated as a soft objective to enhance feasibility under tight geometric constraints. In Phase I, a shrinking-trust-region strategy progressively tightens the soft trust-region radius by increasing the penalty weight, effectively suppressing linearization errors. A sensitivity-driven mesh refinement method then allocates collocation points based on their contribution to the objective function. Phase II applies residual-based refinement to reduce discretization errors. The resulting reference trajectory is tracked using a linear quadratic regulator (LQR) within a reference-trajectory-tracking guidance (RTTG) architecture. Simulation results demonstrate that the proposed method achieves convergence in only a few iterations, generating high-fidelity trajectories within 2–3 s. Compared to pseudospectral solvers, the method achieves over 12× computational speed-up while maintaining kilometer-level accuracy. Monte Carlo tests under uncertainties confirm a 100% success rate, with all constraints satisfied. These results validate the proposed method’s robustness, efficiency, and suitability for onboard real-time entry guidance in dynamic mission environments.