Positioning without GPS.

Every autonomous aircraft assumes it knows where it is. Strip away GPS — through jamming, spoofing, or terrain — and most are lost inside a minute. Inertial navigation buys time, but it drifts: the error grows without bound until something corrects it. In a contested environment, nothing does.

The ground is a map the aircraft can read. A downward camera sees the same terrain a satellite saw before the mission. OtoNav matches the two, frame by frame, and reads its position off the reference — the way a pilot fixes on a river bend or a road junction, but continuously, in software.

Nothing is transmitted. The aircraft is a receiver of its own camera, not an emitter. There is no signal to jam and no beacon to spoof.

Matching a live camera frame to old satellite imagery is where classical computer vision falls over. Seasons change, sensors differ, the map is years out of date — and off-the-shelf feature matchers return nothing.

OtoNav’s matcher is trained for exactly this gap. It finds the fix where the standard methods find zero, in a model small enough to run on the aircraft itself.

A navigation system that reports a confident wrong answer is more dangerous than one that admits uncertainty. OtoNav’s confidence tracks its accuracy closely, so the autopilot can lean on a good fix and coast on inertial through a weak stretch — the same discipline a human navigator uses. It never fakes a lock.