Real-Time Visual Feedback Control of Multi-Camera UAV

Recently, flight control of unmanned aerial vehicles (UAVs) in non-global positioning system (GPS) environments has become increasingly important. In such an environment, visual sensors are important, and their main roles are self-localization and obstacle avoidance. In this paper, the concept of a multi-camera UAV system with multiple cameras attached to the body is proposed to realize high-precision omnidirectional visual recognition, self-localization, and obstacle avoidance simultaneously, and a two-camera UAV is developed as a prototype. The proposed flight control system can switch between visual servoing (VS) for collision avoidance and visual odometry (VO) for self-localization. The feasibility of the proposed control system was verified by conducting flight experiments with the insertion of obstacles.