AI Drone Defeats Human Champions in High-Speed Racing Competition
Engineers at Delft University of Technology developed a lightweight autonomous quadcopter that won an international first-person-view drone race against several human champions. The drone used limited onboard sensors and computing, demonstrating that AI can compete on a physical course designed for human pilots. The team's scientific findings were published in Science Robotics.
Delft University’s Micro Air Vehicle Lab created MonoRace, a lightweight autonomous quadcopter. It competed in Abu Dhabi in November 2025 and beat multiple human FPV champions, a first for an AI drone in an international race. Unlike an earlier Zurich lab victory, this event took place on a course built for human pilots.
FPV racing pilots use goggles linked to a front-facing camera and joysticks to control throttle, roll, pitch, and yaw. Drones pass small gates at 60–100 mph. The Delft team aimed to see how much intelligence could be achieved with modest onboard sensing and processing; findings appeared in Science Robotics.
This result may encourage robotics researchers to prioritize efficient onboard intelligence over heavier sensor and computing packages. It could shape how autonomous drones are designed for tasks where fast decisions and limited resources matter. Human FPV pilots might see new competitive formats or training tools, while race organizers could face questions about separate categories. Wider adoption may depend on demonstrated reliability, safety, and public trust.