深層強化学習を用いたドローンの安定化制御に関する研究

―ドローン用強化学習シミュレーション環境の構築―

抄録

The goal of this research is to realize stabilized control of a drone under serious conditions, especially in strong wind area. We use DDPG (Deep Deterministic Policy Gradient) method, one of the deep-reinforcement learning algorisms which can be used for control in continuous action space. Reinforcement learning is useful for an autonomous tuning of control rules, and deep learning is suitable for obtaining a control mechanism in complex states. In this paper, we explain a proposed control method and simulation environment for reinforcement learning of a drone.