Assess-Ranging-Fuse: Obstacle Detection Based on Stereo Vision Threat Assessment and Ultrasonic Ranging

In this paper, we propose a stability and reliability obstacle avoidance detection system (OADS) for the aggressive flight of autonomous quadrotors. We name our system as assess-ranging-fuse: ARF. In particular, we use the sliding window on the global depth map to determine the nearest obstacle information, instead of directly using the closest depth information, to avoid misjudgment of the environment due to aircraft shaking. Our ARF uses Federal Kalman Filter (FKF) to fusion threat assessment and ultrasound ranging in OADS. Among them, threat assessment by a stereo vision composed of two RGB cameras, the global sliding window to evaluate the threat distance and give the largest object's visual threat distance. Our system is integrated into the quadrotor for indoor/outdoor and dynamic/static experimental tests, and results show that this system can achieve stable global obstacle avoidance.