New AI calculates distant orbits of the planet 100,000 times faster


This site can earn affiliate commissions from the links on this page. Terms of use.

We live in a solar system of eight planets, none of which collide with each other, which is good for us. However, how often do planets in other solar systems collide with each other? A new AI designed by Princeton researchers can analyze the numbers with record speed to determine which potential orbits are stable and which will result in catastrophe. This could help astronomers define the orbits of distant solar systems that we cannot examine in sufficient detail.

Our current exoplanet detection technology cannot provide accurate orbital information, but we can get a general idea of ​​the mechanics by analyzing what we know and modeling the various options. Unfortunately, there are many potential orbits, and modeling a trillion of them can take many hours even with powerful supercomputers. Daniel Tamayo, a fellow with NASA’s Hubble Scholarship Program Sagan Fellow in Astrophysical Sciences at Princeton, devised the algorithm as an alternative to the “brute force” computing that researchers currently use.

Separating potentially stable from unstable orbits is computationally expensive, Tamayo said, even with today’s supercomputers because there are so many orbits to explore. Tamayo’s Stability of Planetary Orbital Configurations Klassifier system simplifies the process by combining a reduced model of planetary interactions with machine learning techniques. This allows SPOCK to quickly rule out the most unstable options, giving you a few thousand plausible orbits in a fraction of a second instead of hours.

At a basic level, the algorithm separates systems that will “break” or break “soon” from stable ones. In this case, “soon” means in the space of a few million years. Given the average lifespan of a solar system, astronomers are unlikely to be seeing any of these doomed configurations. AI begins by simulating 10,000 orbits. SPOCK creates 10 summary metrics from that data to capture the resonant dynamics of the system, and then the algorithm predicts based on these metrics if the settings would remain stable for a billion orbits. This turns out to be approximately 100,000 times faster than traditional methods.

SPOCK can’t tell you exactly what an alien solar system looks like, but it can rule out configurations that are definitely unstable. This could help astronomers narrow their observations as they try to study distant exoplanets. Maybe one day we will have instruments powerful enough to get an accurate picture of the exoplanet’s orbits, but for now, we will have to leave it to the AI.

Now read: