100 km/h without GPS: how does an autonomous racing drone find its way? TU Delft’s MonoRace uses neural gate recognition and learned motor control. Here is what it achieved — and what it could mean for future urgent deliveries.
The demonstrated system used one camera plus an inertial sensor and a learned motor controller running at 500 Hz. It reached up to 100 km/h on a known competition course. This does not demonstrate a medical-delivery service or prove safety around people.
Dates: race 2025; original preprint 21 January 2026; Science Robotics publication 14 October 2026; TU Delft report 16 October 2026. The publication describes earlier experiments.
What would you send above the traffic? Tell us in the comments.
Sources and research imagery (accessed 19 October 2026):
Research photographs/figures: Stavrow A. Bahnam, Robin Ferede, Till M. Blaha, Anton E. Lang, Erin Lucassen, Quentin Missinne, Aderik E. C. Verraest, Christophe De Wagter & Guido C. H. E. de Croon, MonoRace, arXiv:2601.15222v1, CC BY 4.0. Cropped/resized for explanation.
Three brief excerpts from TU Delft/MAVLab’s official film are visual quotations analyzing the demonstrated flight, gate recognition and speed; original soundtrack omitted.
The cyber-city delivery scene and portal cover are AI-generated future illustrations; narration is AI-generated. Our future interpretation is not a statement or endorsement by TU Delft.
#AI #AutonomousDrones #Robotics
source