A team of Dutch researchers has developed a neural network that can look for asteroids that pose a danger to Earth.
Artificial intelligence, developed at the University of Leiden, is called “the ID hazardous objects” (Hazardous Object Identifier — HOI). He identified 11 asteroids that were not previously classified by NASA as a threat. These cosmic bodies are more than 100 meters in diameter and can hit the Ground with a force of hundreds of nuclear warheads. The network also focused on the space objects that can pass within 7.5 million kilometers from Earth. None of the observed asteroids is not yet an imminent threat. However, they have an astronomically small chance of crashing into our planet between 2131 and 2923 years.
With the help of a supercomputer, researchers quickly moved forward with modeling the orbital motions of the planets of the Solar system for 10 thousand years. The team then changed the simulation, simulating the future orbits of asteroids and assessing their ability to approach. This simulation served as the starting material for the neural network, which was looking for patterns in the data common to the simulated asteroids that could eventually hit the Ground. The team compared the network data with data from NASA about known asteroids. Program with accuracy 90,99% was determined which ones were dangerous and which are not. Now scientists are working to make the neural network more accurate.
Video of the day. For believers “curse” of Philip Kirkorov