Artificial intelligence “trained” to search for dangerous asteroids

A group of scientists from the University of Leiden have presented an algorithm capable of detecting potentially dangerous astronomical objects. As reported, at the moment he has already managed to find 11 such celestial bodies.

The first results have already been achieved with its help.

“Potentially dangerous” refers to relatively large space objects, at one stage or another of their movement in outer space capable of being at a distance from the Earth, less than twenty times the average distance from our planet to the Moon. However, all currently known celestial bodies belonging to this category do not really threaten the Earth too much. The probability of a collision with the “most dangerous” of them, the asteroid Bennu, in the period from 2169 to 2199 will be only one in four thousand, most often it is many orders of magnitude less.

Nevertheless, many experts have long been concerned that potentially dangerous asteroids are often detected only shortly before they come close to the planet. To date, it is known about 4,300 near-Earth objects, the diameter of which ranges from one hundred meters to a kilometer, however, according to theoretical calculations, their number can reach 430 thousand, that is, for every known asteroid there are a hundred unknowns. In this regard, scientists from different countries, including Russia, are actively working not only on how to prevent collisions with asteroids, but also on means of detecting celestial bodies approaching the Earth.

The authors of the new study suggested using neural networks for this purpose – the type of computer programs most suitable for the definition of “artificial intelligence” that exist today. Neural networks differ from “ordinary” algorithms in that they are able to “learn” by analyzing a large amount of information and identifying patterns in it, so that they can subsequently be used for independent analysis of similar objects and situations. In this case, the specialists used a supercomputer, which allowed the neural network to analyze the trajectory of many objects in the solar system. This made it possible to establish the possible trajectories of asteroids that could be near the Earth.

The scientific work devoted to the obtained results was published in the journal Astronomy & Astrophysics.

Read the material “Scientists have discovered a darkening of the star Betelgeuse: will turn into a second moon”