We are thrilled to share our groundbreaking paper published today in
@Nature: "Champion-Level Drone Racing using Deep Reinforcement Learning." We introduce "Swift," the first autonomous vision-based drone that beat human world champions in several fair head-to-head races! PDF (open access):
nature.com/articles/s41586-0…. Video:
youtu.be/fBiataDpGIo
The "Swift" AI drone combines deep
#reinforcementlearning in simulation with data collected in the physical world. This marks the first time that an autonomous mobile robot has beaten human champions in a real physical sport designed for and by humans. The key difference to previous achievements, like IBM’s Deep Blue winning against Gary Kasparov at chess in 1996 or Google DeepMind’s AlphaGo crushing the top champion Lee Sedol at Go in 2016, is that the previous feats were limited to board games or video games.
This work represents a milestone for mobile robotics, machine intelligence, and beyond, which may inspire the deployment of hybrid learning-based solutions in other physical systems, such as autonomous vehicles, aircraft, and personal robots, across a broad range of applications.
Kudos to my coauthors
@kaufmann_elia @l_bauersfeld @antoniloq, the pilots
@captainvanover @bmsthomas, and our supporters
@m_schaepman @UZH_en @UZH_Science @UZHspacehub @ERC_Research @nccrrobotics