How to Save a Drone in Case of Low Battery? A New Algorithm for Determining Safe Landing

Low battery is one of the most widespread causes of multi-copter crashes. Most modern drones have straightforward landing algorithms in this type of emergency, which generally decreases altitude until the ground is touched and disarms the motors. If there are wires, fences, or the edges of buildings on the line of descent, this behavior can lead to an accident. This paper proposes a new algorithm based on Random sample consensus and K-Dimensional Tree methods to determine safe landing surfaces around the drone using depth camera data. The algorithm allows not only to avoid obstacles but also to reduce landing time by performing short maneuvers toward the nearest surface suitable for landing. A dataset collected from a real drone was used in experimental studies. They showed that the proposed algorithm provides 5 cm accuracy of a landing point localization from a height of 3 m, making it suitable for small-sized multi-copters.