IROS 2018 Autonomous Drone Race: Optimal Methods meet Deep Learning for Autonomous Drone Racing

IROS 2018 Autonomous Drone Race: Optimal Methods meet Deep Learning for Autonomous Drone Racing
This video shows our team performance at the 3rd edition of the international IROS 2018 Autonomous Drone Racing Competition, which took place during the IROS 2018 conference in Madrid on October 3, 2018, where our team ranked 1st, taking only 30 seconds to complete the track! The competition consisted of navigating autonomously through a number of gates that were moved between laps in order to prevent teams from relying on pure global localization methods (SLAM) or overfitting. The winner was the one would complete the track with minimum time. Our approach was fully vision based and combined machine learning and classical model-based methods. All computation, sensing, and control ran onboard a Qualcomm Snapdragon Flight board and an Intel Up board.

Team participants: Elia Kaufmann, Philipp Foehn, Matthias Gehrig, Julien Kohler, Davide Falanga, Davide Scaramuzza.

Competition webpage: http://ris.skku.edu/iros2018racing/

Our research page on autonomous vision based drone navigation:
http://rpg.ifi.uzh.ch/research_mav.html

Robotics and Perception Group, University of Zurich, 2018
http://rpg.ifi.uzh.ch/

IROS 2018 Autonomous Drone Race: Optimal Methods meet Deep Learning for Autonomous Drone Racing