Autonomous quadrotor obstacle avoidance based on dueling double deep recurrent Q-learning with monocular vision

• A two-stage framework to achieve quadrotor obstacle avoidance with monocular vision is proposed. dueling double deep recurrent Q network trained learn the policy. The employs unsupervised learning based depth estimation for perception. This paper proposes a novel learning-based realize autonomous vision. adopts architecture, consisting of sensing module and decision module. in an manner can extract information from on-board camera image. Moreover, uses Q-learning eliminate adverse effects camera’s limited observation capacity while choosing practical action. has two advantages: (1) it enables without any prior environment or labeled datasets training, (2) its model be easily updated facing new application scenarios. experiments several different simulation scenes show that outperforms high passing rate crowded environments good generalization ability transformed