実機UAVを用いた強化学習によるホバリングの実現

抄録

In this study, reinforcement learning was applied to actual quadrotor unmanned air vehicles (UAVs) to achieve stable hovering. UAVs have various applications, including rescue operations, sports, and entertainment. However, realizing autonomous control of UAVs remains a challenge. Q-learning is a common reinforcement learning method that is used in this study. To date, most of the studies on learning controllers of UAVs have conducted only computer simulations to verify the effectiveness of learning. In contrast, an actual quadrotor UAV has been utilized in this study. The experimental results demonstrate that the UAV could realize stable hovering over a marker placed on the ground.