Space debris control
Dead satellites and fragments are piling up in orbit; collisions create more debris, and there is no way yet to clean up at scale.
Open in the interactive tree →The Kessler syndrome describes a cascade in which collisions create debris that causes more collisions. The 2007 Chinese and 2021 Russian anti-satellite tests and the 2009 Iridium-Cosmos collision added thousands of tracked pieces.
As of October 2026
ESA's Space Environment Report 2026 estimates, for 1 February 2026, about 68,450 objects larger than 10 cm in orbit (roughly 11,300 of them active payloads), 1.5 million debris objects of 1-10 cm and 230 million of 1 mm-1 cm. ESA warns that even with no further launches, collisions among the debris already in orbit would keep the population in low Earth orbit growing. Over the last two decades about 9.8 non-deliberate fragmentations have occurred per year, and continued fragmentations combined with limited disposal success could lead to a cascade of collisions over the next century.
What is missing
- Tracking of objects below 10 cm
- Affordable removal technology (grabbing and de-orbiting large dead rocket bodies)
- Binding international rules and liability for disposal
- Satellites that reliably de-orbit at the end of life
- Traffic coordination among thousands of operators
Becomes possible once solved
- Safe use of low Earth orbit for decades
- Lower collision risk for crewed stations and constellations
- A space economy that does not destroy its own orbits
Open steps
- Tracking objects below 10 cm Medium AI leverageDetect and track millions of 1-10 cm fragments with radar and optical sensors well enough to know where they are.
- Conjunction risk prediction Medium AI leveragePredict which close approaches become real collision risks, to cut false alarms and needless manoeuvres across thousands of satellites.
- Capturing tumbling dead objects Medium AI leverageLet a servicer estimate the motion of a dead rocket body from camera images, then approach and grab it safely without help from the target.
- Long-term debris growth models Medium AI leverageModel how debris grows under different launch, disposal and removal rules to find which action removes most risk per object.
- Reliable end-of-life disposal Low AI leverageRaise the share of satellites that actually de-orbit at end of life, by predicting failures early and designing reliable disposal.
Where AI could help
Low AI leverage. Removal hardware, binding rules and operator cooperation are the bottlenecks; AI helps tracking and collision-risk triage only at the margins.
- Triage conjunction alerts and cut false alarms so operators manoeuvre only when needed
- Detect and track objects below 10 cm in optical and radar data with learned detectors
- Plan autonomous manoeuvres and de-orbit sequences for large constellations
- Model long-term debris growth under different launch and disposal rules
Shown so far
- In August 2020, results of ESA's machine-learning competition on satellite collision-risk prediction showed that only 12 of 97 teams beat a naive 'latest risk' forecast. source
- In August 2025, ESA reported that its CREAM prototype automates conjunction assessment and manoeuvre planning on the ground, though ESA does not describe it as machine learning. source
Prerequisites
Unlocks
- Sustainable Earth orbit2040s?