Robust quadcopter landing on moving ships improved with adversarial training
Curriculum-Based Adversarial Heterogeneous Agent Reinforcement Learning for Autonomous Quad-Copter Landing in Maritime Settings
Machine Learning
Summary
Landing drones on moving ships is hard because of strong winds and a shaking deck. The authors trained a computer system that controls both the drone and a robotic arm on the ship using a special way of learning that includes facing tough wind conditions on purpose. This method helps the system work well even in wind conditions it has never seen before, making drone landings safer and more reliable. Compared to other training methods, this new approach reduces crashes and handles rough seas better.
What this means in practice
- •For drone control engineers: Develop robust autonomous landing systems for drones on ships using adversarial training to improve safety in turbulent maritime conditions.
- •For marine robotics teams: Enhance cooperative control between ship-mounted robotic arms and aerial drones to enable mid-air captures under varying sea states.
Authors
Allan Minh-Tam Nguyen, Sree Showrya Kotala, Stefan Banioi-Crijman, Kurt Driessens, Rico Möckel
Abstract
Recovering unmanned aerial vehicles (UAVs) in maritime environments is challenging due to wind turbulence and ship-deck motion, making it a valuable test case for alternative control and learning approaches as conventional landing approaches often become unreliable. We study simulated mid-air capture of quadrotor UAVs by a ship-mounted robotic arm, learning robust cooperative control policies with Heterogeneous-Agent Proximal Policy Optimization (HAPPO) Reinforcement Learning. We train with HAPPO using a curriculum and an adversarial wind agent (HARL-AC) in NVIDIA Isaac Lab, and compare the obtained control policies against those generated through curriculum-based domain randomization and a benchmark trained on a single sea state. In-distribution evaluation on sea states $0/4/5$ shows comparable success for HARL-AC and domain randomization of up to $97.5\%$. On out-of-distribution sea states $7/8/10$, HARL-AC generalizes better, achieving up to $16\%$ higher median success rate at sea state 10, and substantially lower crash rates of up to $14\%$ compared to the domain randomization policy. Furthermore, we show that the adversarially trained policy shows more cautious behavior, slightly increasing timeouts by $<3\%$, but yields safer recovery behavior in severe, unseen conditions.