06.2 - Flight Dynamics and Control (UAV related)

DECOUPLED DEEP REINFORCEMENT LEARNING FOR AUTONOMOUS LANDING OF UAVS ON SHIPS

T. Gerner¹, M. Candon¹, A.K. Gostar¹, N. Jayarathne, Navantia Australia, Australia; J. Venning, University of Tasmania, Australia; M. Saska, Czech Technical University in Prague, Czech Republic; P. Marzocca¹; ¹RMIT University, Australia

Autonomous UAV landing on ship decks is challenged by stochastic winds and wave-induced motion. This paper presents a decoupled actor–critic reinforcement learning framework with separate horizontal/attitude and vertical controllers. Evaluated in a Gazebo–ROS–PX4 simulation, the approach achieves robust deck tracking with 25 cm horizontal error and soft landings with 0.35 m/s RMS relative velocity under varying wind conditions.

View Paper

DOI: 10.71945/icas2026_0867