Dynamic Obstacle Avoidance of Quadcopters with Monocular Camera Based on Image-based Visual Servo

This paper focuses on avoiding a dynamic obstacle for Micro-aerial vehicles (MAVs). Instead of estimating the depth or the 3D model of the obstacle, our method is only based on images. First, a geometric model is established, which describes the relationship between a quadcopter and a dynamic obstacle by a monocular camera. Then, based on the established geometric model and the dynamics of the quadcopter, the avoidance control command, namely the angular velocity and the thrust, are generated in real-time. Finally, both the hardware-in-the-loop simulation and the flight experiment are demonstrated to verify the effectiveness of the algorithm.