Autonomous navigation, landing and recharge of a quadrotor using artificial vision

In this paper a solution to UAV reduced endurance and autonomous flight is proposed. With a complete on-board solution, based on artificial vision, the developed system is able to autonomously take off, navigate and land, recharging its battery by using a dedicated landing platform, both in indoor and outdoor scenarios. The landing platform includes a passive centering system to correct the landing error of the UAV, with a novel design wich reduce cost and increase the safety (thanks to small and isolated electrical contacts) without invasive hardware changes on the drone. The developed vision algorithm provides a fast and accurate measurement of UAV position with respect to the landing platform using a visual target, but at the same time it is able to automatically switch to an estimation of position that is independent from the visual target. This aspect is used during navigation or when the tracking of the target fails, ensuring a continuous position measurement feed to the controllers. The developed control system manages all the different phases of a mission (motor turning on/off, take off, navigation, landing, ...) with low control error, ensuring a landing over the landing platform with an error that is lower than 5cm for both x and y axis. The developed software in ROS environment is modular and provides input/output interfaces to receive command, or send data.