AI drone wins race without GPS
During a competition in Abu Dhabi, an AI-piloted racing drone named MonoRace defeated all its autonomous competitors and three former FPV drone racing world champions, achieving speeds exceeding 100 km/h. The AI did not use traditional sensors like GPS, preloaded track maps, or stereoscopic cameras, relying instead on a monocular camera and an inertial measurement unit (IMU) to track its position and movement in real time.MonoRace's unique navigation method
Developed by researchers from the micro-air vehicle laboratory at Delft University of Technology in the Netherlands, MonoRace used an innovative approach to estimate obstacles and circuit gates. The AI identified start gates and calculated the drone’s position and movement to generate appropriate motor commands, which were sent to the four motors 500 times per second via the drone’s flight controller.AI mimics human pilot behavior
Researchers from the University of Zurich redesigned AI algorithms to make the drone behave like a real pilot, overcoming challenges such as camera interference and IMU saturation. Despite these limitations, the drone performed at a champion level, showing that AI can make high-speed piloting decisions based on limited information similar to that of human pilots.Implications for autonomous systems
The success of MonoRace has broader implications beyond drone racing. The AI’s ability to perceive and react in complex environments without GPS could improve the autonomy of flying robots in situations where GPS is unavailable or unreliable. With the war in Ukraine highlighting the increasing use of drones and AI in military applications, the development of such systems raises questions about future technological progress in both civilian and military contexts.
