Trajectory Reconstruction of Blunt Body Capsules from Ballistic Range Testing Data Using Multi-Fidelity Gaussian Processes

Trajectory reconstruction from the sparse observation points of ballistic range testing data is crucial for accurately evaluating the performance and behavior of typical blunt bodies like the Genesis entry capsule, which become dynamically unstable at mid to low supersonic speeds. This instability causes the angle of attack oscillations during descent; if not addressed during the design process, these oscillations may lead to mission failure. Traditional methods often struggle with uncertainties and noise in the measurement data. Gaussian process regression offers a robust, non-parametric approach to address these challenges. However, the Gaussian process based on classical Bayesian optimization strategies does not effectively accommodate the inclusion of known physical behaviors or prior knowledge. This paper explores a hybrid optimization technique incorporating free-flight computational fluid dynamic simulations as prior information in physical models into the Bayesian optimization framework. We investigate applying this Multi-Fidelity Gaussian process for trajectory reconstruction in synthetic ballistics testing data. We emphasize its advantages in handling noisy data and providing probabilistic predictions, which can be used as input to parameter identification tools. The results indicate that the framework improves the precision of ballistic trajectory predictions and offers robust uncertainty quantification, making it an invaluable tool for ballistic range data analysis and design optimization.