Landing Footprint Prediction for Hypersonic Morphing Vehicle: A Generative Adversarial Network-Based Method

Hypersonic morphing vehicles (HMVs) require multiple landing footprint predictions for decision-making and target allocation. Unlike the traditional hypersonic vehicle, the varying aerodynamic characteristics corresponding to different shapes pose challenges to the landing footprint prediction. This study proposes a rapid landing footprint prediction method for HMV based on the dual-generator Wasserstein conditional generative adversarial network (DG-WCGAN). By leveraging the learning capabilities and computational efficiency of generative artificial intelligence, this method eliminates integral calculations and accurately predicts the landing footprint for the current shape or morphing-enabled situation. First, an aerodynamic parameters and landing footprint database is constructed using aerodynamic simulation software and numerical optimization. Further, the structure of the generative adversarial network is tailored by incorporating a dual generator. This enables the one-step computation of aerodynamic parameters and landing footprint and offers the flexibility to adjust the number of generated landing points according to the requirements. Moreover, the extra network can be independently utilized for aerodynamic parameter calculations in HMV control. Numerical simulations demonstrate that the proposed DG-WCGAN method improves computational speed by 21.7% while reducing prediction error by 68.6% compared to the deep neural network approaches and exhibits better generalization capability.