What Counts as a Close Approach in Orbit
Satellite operators call a predicted close pass between two orbiting objects a conjunction. It is not a collision, only a forecast that two paths in space will come near each other at a specific future moment, based on tracked orbital data.
With tens of thousands of satellites and debris fragments now catalogued, conjunctions happen constantly. Most predicted close approaches are harmless once uncertainty narrows, but each one still has to be screened before it can be ruled out as a real risk.
Who Tracks Every Object in Earth Orbit
The U.S. Space Force runs a global network of radars and telescopes, the Space Surveillance Network, that tracks tens of thousands of objects larger than about ten centimeters in low Earth orbit and reports their positions publicly.
This tracking data feeds catalogs like Space-Track.org and the open-source CelesTrak, which operators, researchers, and even hobbyists use to check whether their own satellites are on a path to pass close to something else.
The Conjunction Data Message Explained
When tracking data flags a potential close approach, the Space Force or a partner agency issues a Conjunction Data Message, a standardized report listing the two objects, the predicted miss distance, the time of closest approach, and the uncertainty in each orbit.
Operators feed these messages into their own screening software, combining the official tracking data with their satellite's more precise onboard navigation data to decide whether the predicted pass is actually worth acting on.
Calculating the Probability of Collision
Rather than treating every conjunction as equally dangerous, operators calculate a probability of collision, a single number that combines the predicted miss distance with the physical sizes of both objects and the statistical uncertainty in each tracked orbit.
A tight predicted miss distance with large orbital uncertainty can produce a higher collision probability than a wider miss with well-known orbits, which is why raw distance alone is never used as the deciding factor.
Why Miss Distance Alone Isn't Enough
Two objects predicted to pass within a few hundred meters of each other might carry very different risk depending on how precisely their orbits are known. Older tracking data or infrequent updates widen that uncertainty and push the risk estimate up.
This is why operators wait as long as they safely can before deciding to maneuver: newer tracking data collected closer to the predicted approach time narrows the uncertainty and often lowers the calculated probability of collision on its own.
The Industry-Standard Risk Threshold
A widely used benchmark across the satellite industry sets the action threshold at a collision probability of roughly one in ten thousand. Below that number, most operators treat a conjunction as background noise, not something requiring a maneuver.
The European Space Agency has historically used this same one-in-ten-thousand threshold for its own low-Earth-orbit missions, executing an avoidance burn only when a conjunction's calculated risk crosses that line during final screening.
Starlink's Tighter Safety Margin
SpaceX has described Starlink's own action threshold as considerably stricter than the industry norm, moving a satellite well before its calculated collision probability would reach the one-in-ten-thousand line other operators use as their cutoff.
SpaceX and independent trackers have reported this stricter threshold at roughly one collision in every one hundred thousand conjunctions, a margin that trades a much higher number of precautionary maneuvers for a much lower chance of ever being wrong.
Handing the Decision to an Onboard Computer
Starlink satellites run an autonomous collision-avoidance system that ingests public conjunction data and the company's own tracking, then decides on its own whether a maneuver is needed and calculates the exact burn without waiting for a human operator.
The satellite computes the size, direction, and timing of the avoidance burn itself, executes it, and returns to its assigned orbital slot afterward, all inside a decision loop that runs continuously across the entire fleet.
Why No Human Can Review 10,000 Satellites
SpaceX has stated plainly that autonomous maneuvering is the only practical way to operate a fleet of more than ten thousand active satellites. No team of human operators could review every predicted conjunction fast enough at that scale.
A single satellite might face dozens of flagged conjunctions in a day as tracking data updates, most of them resolving as harmless once newer data arrives. Automating that filtering step is what makes megaconstellations operationally possible at all.
How Many Maneuvers Happen Each Day
Reporting on Starlink's own safety data shows the constellation accumulated more than 355,000 collision-avoidance maneuvers between June 2025 and May 2026, averaging around 822 automated avoidance burns across the fleet every single day.
That figure has climbed sharply as SpaceX has added satellites and lowered its risk tolerance further, illustrating how quickly maneuver volume scales with constellation size even when each individual maneuver only shifts a satellite's orbit slightly.
What a Collision-Avoidance Burn Looks Like
Most avoidance maneuvers are small nudges, not dramatic swerves. A satellite typically raises or lowers its orbit by a modest amount for a limited window, enough to shift its position by the safety margin needed at the predicted approach time.
Once the risky window has passed, the satellite generally returns to its assigned altitude and orbital slot so it doesn't drift out of formation with the rest of the constellation or create a new conjunction risk elsewhere.
The Thrusters That Make the Nudge Possible
Modern constellation satellites like Starlink use electric ion thrusters, which produce gentle but continuous thrust using electrically charged propellant instead of a chemical rocket engine's brief, forceful burn.
This kind of propulsion is well suited to frequent small adjustments over a satellite's operational life, since it uses very little propellant per maneuver, letting a satellite perform many avoidance burns without exhausting its fuel supply early.
Retiring a Satellite Safely
When a satellite reaches the end of its working life, operators typically lower it into a decaying orbit so atmospheric drag pulls it down and burns it up on reentry within a bounded time frame, clearing it from the active traffic picture.
U.S. regulators have moved to tighten how quickly this has to happen. Newer orbital debris rules push operators of low-Earth-orbit satellites toward deorbiting within five years of the end of a mission rather than leaving dead hardware drifting for decades.
The Kessler Syndrome Risk
The Kessler syndrome is a scenario, first described by NASA scientist Donald Kessler in 1978, in which a collision creates debris that goes on to trigger further collisions, cascading into a growing debris field in low Earth orbit.
Collision avoidance between active satellites is one of the main defenses against this scenario, since preventing a first collision also prevents the fragmentation event that could set off a chain reaction of further debris-generating impacts.
What's Actually Floating Up There
The tracked object population in orbit includes active satellites, dead satellites that were never deorbited, spent upper stages from rocket launches, and fragments left over from past collisions, explosions, and intentional weapons tests.
Only a small fraction of this population is functioning spacecraft. The majority is inert hardware and fragments, which is exactly why conjunction screening has to check every satellite against the entire catalog, not just against other active missions.
Ground-Based Radar and Optical Tracking
The Space Surveillance Network combines ground-based radar, which can detect objects even through cloud cover, with optical telescopes that photograph reflected sunlight off objects at higher altitudes where radar is less effective.
Each sensor contributes position measurements that get combined into a refined orbit for every catalogued object, and that refined orbit is what actually gets compared against other objects' orbits to flag potential conjunctions.
How Many Objects Are on the Public Catalog
The publicly tracked catalog run by the U.S. Space Force lists tens of thousands of objects in orbit, a number that has grown sharply over the past decade as launch rates have accelerated and sensor coverage has improved.
Sensor improvements mean smaller and smaller objects keep getting added to the catalog over time, so the count reflects both a genuinely busier orbital environment and a steadily improving ability to actually see what's already there.
The Debris You Can't See Coming
Objects smaller than about ten centimeters are generally too small for the Space Surveillance Network to reliably track, yet even a paint fleck or bolt moving at orbital speed carries enough energy to seriously damage a satellite.
This untracked population is a real limitation on conjunction screening. Operators can only avoid what they know is coming, so uncatalogued debris remains a residual risk that maneuvering strategies can reduce but never fully eliminate.
Data-Sharing Agreements Between Operators
Beyond public tracking data, major operators increasingly exchange orbital information directly with each other. SpaceX and NASA, for example, established a coordination agreement covering Starlink's conjunctions with NASA science and crewed spacecraft.
These bilateral arrangements let two operators compare their own higher-precision navigation data rather than relying solely on the lower-precision public catalog, producing sharper risk estimates and clearer agreement on who maneuvers when a conjunction is flagged.
How Many Maneuvers ESA Used to Fly
Before megaconstellations reshaped low Earth orbit, the European Space Agency reported executing an average of around twelve collision-avoidance maneuvers per year across its low-Earth-orbit fleet, a manageable number for human-reviewed operations.
That historical figure is a useful baseline for how much the environment has changed. A single constellation now performs hundreds of automated avoidance burns per day, a scale entirely different from the human-in-the-loop era that figure describes.
The Night ESA Dodged a Starlink Satellite
In 2019, the European Space Agency maneuvered its Aeolus satellite to avoid a predicted close approach with a Starlink satellite, an early, widely reported example of a megaconstellation forcing another operator to take action.
ESA later noted that coordination around that event was slower and more manual than it wanted, and cited it as a driver toward automating conjunction screening and improving direct communication channels between operators of different constellations.
Conjunction Rates Are Climbing With Every Launch
Every new satellite added to orbit increases the number of possible pairings with everything already up there, so conjunction rates rise faster than the number of satellites itself as constellations grow.
This is one reason space agencies and operators describe current tracking and screening practices as adequate for today's traffic but under real strain, with planned constellations from several countries and companies expected to add tens of thousands more satellites.
What Space Traffic Management Actually Means
Space traffic management is the broader effort to standardize how conjunctions are calculated, shared, and acted on across every operator, rather than leaving each company or agency to run its own separate process.
It covers things like common data formats, agreed maneuver responsibility rules, and shared screening thresholds, aiming for something closer to how air traffic control coordinates aircraft, though space traffic management is far less mature than its aviation counterpart.
Why No Single Authority Runs Space Traffic
Unlike aviation, there is no single international body with binding authority over every satellite operator's maneuver decisions. National regulators, industry norms, and voluntary agreements fill that gap unevenly across different countries and companies.
This patchwork works reasonably well today because a handful of large operators account for most active satellites and generally follow similar thresholds, but it leaves real gaps as more countries and private companies launch their own constellations.
The Software Behind Conjunction Screening
Automated screening tools continuously compare every catalogued orbit against every other, flagging pairs whose predicted paths will come within a defined distance of each other over a coming window, typically a week or less.
Open resources like CelesTrak publish screening tools and orbital element sets that let smaller operators, researchers, and even amateur satellite trackers run their own conjunction checks without needing access to classified or proprietary tracking systems.
Machine Learning Enters the Collision-Avoidance Loop
The European Space Agency has run public machine-learning competitions aimed at predicting collision risk more accurately from raw tracking data, part of a broader push toward the automated conjunction screening that megaconstellations now require.
A related ESA proposal, known as CREAM, aims to automate not just risk estimation but also the decision of whether and how to maneuver, moving the entire process closer to the fully autonomous approach SpaceX already runs on Starlink.
Commercial Companies Selling Space Situational Awareness
Beyond government tracking, a growing number of private companies now sell what's called space situational awareness, using their own ground-based radar networks to track objects and sell refined conjunction data to satellite operators.
These commercial services often track smaller objects or update orbits more frequently than the public catalog, giving paying operators an extra layer of tracking precision on top of the freely available government data.
When Two Satellites Actually Collided
In February 2009, the defunct Russian satellite Cosmos 2251 collided with the active Iridium 33 communications satellite over Siberia, destroying both and scattering debris that is still being tracked today, more than fifteen years later.
It remains one of the clearest real-world proofs that conjunction screening isn't a theoretical exercise. The collision happened because the predicted conjunction wasn't flagged as high-risk enough to trigger a maneuver by either side beforehand.
De-Conflicting Orbits Before Launch
Collision avoidance doesn't start once a satellite is already in orbit. Regulators require constellation operators to submit orbital plans in advance, letting agencies check new launches against existing satellites before they ever leave the ground.
This early-stage screening reduces, but doesn't eliminate, the number of conjunctions that show up later, since actual orbits always drift slightly from filed plans and other operators are simultaneously adjusting their own constellations too.
Regulatory Rules for New Constellations
The U.S. Federal Communications Commission adopted updated orbital debris mitigation rules in 2022, tightening requirements around how quickly newly licensed low-Earth-orbit satellites must deorbit once their working life ends.
Regulators frame these rules as a preventive measure: fewer dead satellites lingering in busy orbital shells means fewer objects that active satellites, and the automated systems watching over them, have to screen against for decades after launch.
Why Automation Is the Only Path Forward
Every trend in this system points the same direction: more satellites, more conjunctions, and less time to review each one manually. Automation isn't a convenience here so much as a structural requirement for constellations of this size to function at all.
That's why SpaceX's autonomous system, ESA's push toward machine learning, and industry-wide efforts at shared screening standards are all converging on the same goal: a system where computers, not people, make the vast majority of real-time avoidance calls.
What Better Space Traffic Control Could Look Like
Proposals for the future include shared, real-time data exchanges between all major operators rather than one-off bilateral agreements, standardized maneuver-responsibility rules so both sides in a conjunction always know who is expected to move, and wider use of machine learning to cut false alarms.
None of this replaces the basic physics of the problem: more objects in a finite volume of useful orbital space means more conjunctions no matter how good the software gets, which keeps collision avoidance a permanent, growing part of running any satellite constellation.
Sources
- Starlink's collision-avoidance maneuvers and space safety data
- ESA: Automating collision avoidance
- Wikipedia: Kessler syndrome
FAQ
What is a conjunction in satellite terms?
A conjunction is a predicted close approach between two orbiting objects at a specific future time, calculated from their tracked orbits. It is a forecast, not an actual collision.
How does SpaceX know when to move a Starlink satellite?
Each Starlink satellite runs an autonomous system that ingests public conjunction data and SpaceX's own tracking, calculates a collision probability, and decides on its own whether and how to maneuver.
What collision probability triggers a Starlink maneuver?
SpaceX has described its threshold as far stricter than the roughly one-in-ten-thousand industry norm, with reports putting Starlink's own action threshold at around one in one hundred thousand.
How many collision-avoidance maneuvers does Starlink perform?
Reporting on SpaceX's own safety data shows more than 355,000 avoidance maneuvers between June 2025 and May 2026, averaging around 822 automated burns across the fleet per day.
Who tracks space debris and satellites?
The U.S. Space Force's Space Surveillance Network tracks tens of thousands of objects using ground-based radar and optical telescopes, publishing data through catalogs like Space-Track.org and CelesTrak.
What is a Conjunction Data Message?
It is a standardized report listing two objects predicted to pass close to each other, their predicted miss distance, the time of closest approach, and the uncertainty in each tracked orbit.
What is the Kessler syndrome?
It is a scenario, first described by NASA scientist Donald Kessler in 1978, where a collision creates debris that triggers further collisions, cascading into a growing debris field in orbit.
Did two satellites ever actually collide?
Yes. In February 2009, the defunct Cosmos 2251 collided with the active Iridium 33 satellite over Siberia, destroying both and creating debris still tracked today.
How does ESA decide when to move its satellites?
ESA has historically used a collision probability threshold of roughly one in ten thousand for its low-Earth-orbit missions, executing an avoidance burn once a conjunction's calculated risk crosses that line.
Can satellite operators talk to each other directly?
Yes, increasingly. SpaceX and NASA, for example, have a coordination agreement covering Starlink's conjunctions with NASA spacecraft, exchanging higher-precision navigation data directly rather than relying only on public catalogs.
What happens to satellites when they retire?
Operators typically lower a retired satellite into a decaying orbit so atmospheric drag pulls it down and burns it up on reentry, often within a regulated time window such as five years.
Why can't humans review every close approach?
With tens of thousands of tracked objects and constellations of over ten thousand satellites, the volume of predicted conjunctions vastly exceeds what any human team could review in real time.
What's the difference between debris and an active satellite in tracking terms?
Tracking systems treat both the same way, as catalogued objects with an orbit, but active satellites can maneuver to avoid a conjunction while debris cannot, making debris a passive risk that only the other side can dodge.
Is there a global space traffic control authority?
No single international body has binding authority over every operator's maneuver decisions. National regulators, industry norms, and voluntary bilateral agreements fill that gap today.
What role does machine learning play in collision avoidance?
ESA has run public machine-learning competitions to improve collision-risk prediction accuracy, and proposals like its CREAM concept aim to automate maneuver decisions further, building on the fully autonomous approach SpaceX already uses.
About the Author
We reference Wikipedia and other authoritative sources to explain the background and current understanding of this topic.
Loved This Article?
Share it on WhatsApp β Share it on WhatsApp
Get more guides in your inbox β Subscribe to our newsletter for weekly surprising stories from Egypt, Saudi Arabia, Dubai, and beyond.