AI learns to reconstruct precise GPS routes from cell tower signals

Researchers developed a video-generation model that can infer detailed GPS trajectories from coarse cellular signaling data that phone carriers already collect. This makes it possible to reconstruct where someone was moving with much higher precision than cell tower IDs alone would allow, using only the signals phones constantly exchange with networks.

Phone carriers have always recorded which cell towers phones connect to, but those records are too coarse to show actual movement paths — this work demonstrates that an AI trained on map images can fill in the missing GPS details, which means cellular data that was previously too grainy to track movement now becomes usable for that purpose at scale.