Space around Earth is increasingly congested. According to the latest data, over 45,000 artificial objects—from active satellites to debris from old launches—are orbiting the planet. With the growth of low-Earth orbit mega-constellations like Starlink, OneWeb, and dozens of new missions planned in 2026, the risk of collisions is becoming a major concern.

To better understand orbital trajectories and collision risks, scientists at Lawrence Livermore National Laboratory (LLNL, USA) created a massive simulation: they calculated one million orbits in cislunar space (from geostationary orbit to beyond the Moon) over a six-year period.

What the LLNL Team Did

The team used the open-source Space Situational Awareness Python package (SSAPy), public orbital databases, and the lab’s Quartz and Ruby supercomputers. Calculating one million trajectories required 1.6 million CPU-hours—on a regular computer, this would take over 182 years, but on the LLNL cluster it was completed in just three days.

The results are openly available at https://gdo-cislunar.llnl.gov/. This isn’t just a map; it’s a high-precision trajectory dataset that can be used for machine learning, statistical analysis, and orbital prediction.

LLNL scientists Denver Higgins and Travis Yeager highlighted the benefits:

“With a million orbits, you can do very rich analysis with machine learning. You can predict orbital lifetimes, stability, or detect anomalies—see if an orbit starts behaving strangely,” said Higgins.
“If you want to know where a satellite will be next week, there’s no simple equation. You have to step forward incrementally,” added Yeager.

Key Findings from the Simulation

  • About half of the simulated orbits remained stable for at least one year.
  • Fewer than 10% stayed stable throughout all six years.

These results help identify “crossroads” zones, where orbital paths frequently intersect and the risk of collisions is highest.

Such data is crucial for:

  • Planning new missions and satellite placement to avoid conflicts
  • Predicting how small perturbations (like solar wind or lunar gravity) affect orbits
  • Monitoring space traffic in real time

Why This Matters in 2026

A record number of launches is expected in 2026, mostly to low-Earth orbit (LEO), which is already crowded. Without global coordination—which currently doesn’t exist—every new satellite increases the chance of a Kessler cascade.

The LLNL dataset is open and free, providing a foundation for countries, companies, and agencies to better coordinate orbital paths. While it’s not yet a real-time collision avoidance system like Space-Track or LeoLabs, it lays the groundwork for future machine learning models that could warn of dangerous close approaches.

In Brief

LLNL simulated one million cislunar orbits over six years, producing an open dataset for predicting orbital stability, lifespan, and collision hotspots. Half of the trajectories remain stable at least a year, less than 10% for all six years. With more than 45,000 objects in orbit, this resource could become key to safer space traffic management in 2026 and beyond.