The research aims to develop a robust control algorithm for helicopter landings on ships, addressing challenges from model uncertainty and environmental disturbances.
Developed a hybrid control algorithm integrating a model-based controller and residual reinforcement learning.
Utilized a split-channel incremental nonlinear dynamic inversion outer loop and a reduced-order dynamic inversion inner loop for the baseline control pathway.
Performed simulations comparing the proposed Residual PPO against baseline controllers and Pure PPO.
Residual PPO improved hover robustness and landing performance compared to the baseline controller and Pure PPO.
Achieved approximately 20–30% residual authority with a 90.0% desired landing rate across various descent-and-landing scenarios.