We are happy to release the code of our Science Robotics paper “Learning
High-Speed Flight in the Wild”.

The code allows you to train end-to-end navigation policies to navigate
drones (or any other mobile robots) at very high speeds in previously
unknown, challenging environments (snowy terrains, derailed trains,
ruins, thick vegetation, and collapsed buildings), with only onboard
sensing and computation. Training is done exclusively in simulation with
imitation learning from a privileged expert. Thanks to a sensor
abstraction procedure, the policies trained in simulation can be applied
to a real platform without any fine-tuning on real data!

All the code and datasets to reproduce the experiments you see in the
video are released for free to the public. Paper, code, dataset, video:
http://rpg.ifi.uzh.ch/AgileAutonomy.html

Antonio Loquercio, Elia Kaufmann, Rene Ranflt, Matthias Mueller, Vladlen
Koltun, Davide Scaramuzza