Vision-Based Autonomous Drone Landing on Moving Platforms via Deep Reinforcement Learning

This video presents our work, “Vision-Based Autonomous Drone Landing on Moving Platforms With Uncertain Motion via Deep Reinforcement Learning,” accepted to IEEE Robotics and Automation Letters (RA-L) 2026. We propose a vision-based deep reinforcement learning framework for autonomous quadrotor landing on moving platforms with uncertain and accelerating motion. The framework jointly learns robust relative state estimation and active perception, enabling the drone to infer the platform’s motion from onboard sensing while choosing actions that preserve reliable visual information during landing. The video includes: 1. Overview of the proposed vision-based DRL landing framework 2. Simulation results under diverse platform motions, including linear acceleration, zig-zag, and boat-like motion 3. Real-world experiments on a moving landing platform using onboard vision and computation Authors: Woojae Shin, Minwoo Kim, Taewook Park, Geunsik Bae, Seunghwan Kim, and Hyondong Oh