The cameras of a self-driving car capture light like telescopes aimed at a distant galaxy. Each frame is a fragmented observation; together they must form a map of the road ahead.
The computer, like an astronomer, doesn't just stitch images together: it generates several possible bird's-eye views and compares them to real data. The best option is chosen based on entropy — a measure of uncertainty that drops when the draft matches reality. Individual frames are processed as in photometry, where brightness and shadows help reconstruct an object's appearance. And the 'observing at different times' approach is borrowed directly from the transit method of hunting exoplanets: there, scientists spot worlds by the tiniest fluctuations in a star's light.
In the end, the driverless car gets a clear blueprint that fears neither rain nor dirt. Tests confirmed: the method is more accurate than previous ones at finding lanes and plotting a safe route.
In nature, the 'bird's-eye view' is used by birds of prey: a peregrine falcon spots prey from the height of a skyscraper because its brain instantly fuses the image from two eyes into a three-dimensional picture.
In the movie 'Ex Machina,' the android behind the wheel relies only on sensors without building a map. Today's algorithms do exactly that — turn scattered signals into a complete scheme for mass-market cars.