Reinforcement-Learning Control of a Hybrid Airship Using a High-Fidelity Digital Twin
A reproducible high-fidelity digital twin is developed for a custom hybrid airship, combining 6-DoF rigid-body dynamics with hull aerodynamics, rotor thrust, and actuator dynamics. Within this virtual testbed, a reinforcement-learning autopilot is trained to steer from arbitrary initial conditions to specified waypoints. Comparative experiments show <3.5 m RMS position-tracking error across multiple unseen missions.