Colloquium: Jacopo Carradori (Space Flight)

Landing a rocket on Earth is a key factor in enabling quicker and more cost-effective access to space. However, it poses significant challenges due to the dynamic and highly uncertain environment. A robust Guidance, Navigation, and Control (GNC) system is essential to guide the vehicle to the landing site while meeting terminal constraints and minimizing fuel consumption. This research integrates Meta-Reinforcement Learning with Gated Transformer XL neural networks to enhance the robustness of terminal powered guidance to atmospheric and aerodynamic uncertainties, navigation and control errors, and dispersed initial conditions. By employing a 6-degrees-of-freedom (DOF) dynamics model and more accurate vehicle and environmental simulations, the agent learns a higher fidelity guidance policy compared to existing literature, demonstrating successful and robust performance in Monte Carlo simulations. In this complex scenario, the innovative attention-based neural networks outperform the widely used recurrent neural networks, typically considered state-of-the-art for Reinforcement Learning-based space guidance applications.

Supervisors: Erwin Mooij