AI drone outpaces former world champions using one camera and no GPS
Aerospace EngineeringRobotics
THE AI ANGLE
Directly generating motor commands from sensor inputsResearchers at TU Delft developed an autonomous drone that outpaced former world champion pilots using one forward-facing camera and an inertial sensor without GPS. A compact neural network runs directly on an onboard chip to issue motor commands 500 times per second. This demonstrates that autonomous aerial robots can achieve high-speed navigation with minimal hardware and computing resources.
THE TEACHING ANGLE
Engineers must weigh direct neural-network motor control against traditional multi-layered controllers that plan explicit paths.Read the original at techxplore.com Generate teaching or study materials
More in Aerospace Engineering
- Royal Navy goes fishing for suppliers to build its Type 93 robot submarineThe Register · October 15, 2026
- Agility chooses Nyobolt for ultra-fast charging for Digit 5 humanoidThe Robot Report · October 15, 2026
- Drones on the line: New algorithms help drone fleets inspect power infrastructurePhys.org — Technology · October 15, 2026
- Soft robotic fish combines high-speed swimming with 3D maneuverabilityPhys.org — Technology · October 14, 2026
- Q&A: Mars Perseverance Rover Sets a Distance RecordIEEE Spectrum — Robotics · October 14, 2026