Autonomous System for a Racing Quadcopter

In this paper, we present a methodology to make an autonomous drone fly through a sequence of gates only with on-board sensors. Our work is a solution to the AlphaPilot Challenge, proposed by the Lookheed Martin Company and the Drone Racing League. First, we propose a strategy to generate a smooth trajectory that passes through the gates. Then, we develop a localization system, which merges image data from an on-board camera with IMU data. Finally, we present an artificial vector field based strategy used to control the quadcopter. Our results are validated with simulations in the official simulator of the competition and with preliminary experiments with a real drone.