Urban concrete jungles turn drones into blinded bats. Cameras go blind around every corner, GPS gets lost among skyscrapers, and radio waves scatter into a hundred echoes. The M3F-UAV model overcomes this by merging all available 'ears' and 'eyes': regular cameras, 3D maps from a laser rangefinder, and radio signals — their electromagnetic nature was first described by Maxwell. The AI is trained on entropy, meaning the inevitable disorder in incoming data. Just as a bat learns to isolate prey from acoustic chaos, the algorithm was forced to reconstruct the scene from deliberately damaged sensor streams. If any part fails — say, the camera is blinded by sunlight — the system doesn't stop. It fills in reality from memory and hints, much like spectroscopy extracts a star's secrets from its rainbow spectrum.
The most striking experiment: a drone with cameras and navigation disabled flew half a kilometer through a model city, navigating solely by radio echoes from buildings — pure echolocation, like cave-dwelling bats. Tomorrow, swarms of such drones could search for people under rubble, operate in thick fog and dust storms, delivering cargo without fear of losing sight.
The first autopilots for drone torpedoes appeared in 1917: they kept course using a gyroscope.
In Terminator 2, the HK-Aerial hunter drone combined infrared vision and radar — anticipating today's multisensory approach.