Researchers and government agencies are turning to artificial intelligence to solve one of spaceflight’s fastest-growing problems: tracking debris in the vast stretch between Earth and the Moon. With more than 100 lunar missions planned for the coming decade, the region known as cislunar space is about to get far more crowded, and current tracking systems were never designed to monitor objects at those distances. A wave of new AI-driven tools, open datasets, and sensor experiments now aims to close that gap before collisions put future missions at risk.
Why Cislunar Space Is a Blind Spot
Most space surveillance infrastructure focuses on low Earth orbit, where the U.S. Space Force’s 18th Space Defense Squadron maintains the public catalog used for collision warnings. That catalog and its conjunction assessments build on a legacy of identification and tracking systems originally optimized for satellites only a few hundred to a few thousand kilometers above Earth. Beyond that range, radar returns weaken, optical targets grow faint, and orbits become harder to predict.
Cislunar space, the zone stretching roughly 266,000 kilometers from Earth toward and beyond the Moon, sits largely outside the reach of those near-Earth sensors. The geometry alone is daunting: objects can appear close to the Sun from our vantage point, vanish behind the Moon, or move slowly enough across the sky that they blend into background stars. Traditional tracking methods that work well for low Earth orbit quickly run into physical and computational limits when extended to this far larger volume.
The scale of the problem is about to grow sharply. According to analysis from the Center for Responsible AI at the University of Arizona, more than 100 lunar missions are planned for the coming decade. Each mission can leave behind spent rocket stages, adapters, and other hardware. Over time, those remnants accumulate in loosely bound orbits influenced by both Earth and lunar gravity. Without a reliable way to catalog them, mission planners face rising uncertainty about what is already out there and where it is headed.
AI-Powered Visibility Maps Fill the Gap
One of the most promising approaches comes from Purdue University, where researcher Carola Frueh and her students are building what they call “visibility maps” for cislunar space. These maps use computational models to predict where telescopes can actually detect debris, accounting for lighting conditions, orbital geometry, and sensor limitations. The goal is to remove blind spots by showing operators exactly which parts of cislunar space are observable at any given time and which remain hidden.
This matters because traditional tracking relies on repeated observations to refine an object’s orbit. In cislunar space, gravitational influences from both Earth and the Moon create chaotic trajectories that are far harder to predict than the relatively stable orbits closer to Earth. If a telescope cannot see an object during a critical window, the uncertainty in its predicted path balloons quickly. Frueh’s visibility maps give ground stations a way to prioritize where to look, turning a nearly impossible search problem into a targeted one and making better use of scarce telescope time.
The same modeling techniques could eventually extend beyond debris. As agencies expand exploration of the solar system, they will need to track crewed vehicles, logistics craft, and scientific platforms moving through cislunar space and on toward more distant destinations. AI-optimized visibility planning offers a scalable way to manage that complexity.
A Million Simulated Orbits for Machine Learning
AI models are only as good as the data they train on, and cislunar tracking has historically suffered from a severe data shortage. A recently published benchmark dataset on arXiv addresses that directly. The dataset, produced by contributors affiliated with Lawrence Livermore National Laboratory’s SSAPy ecosystem, contains one million numerically propagated cislunar trajectories stored in CSV and HDF5 formats with accompanying metadata. Each trajectory is propagated for up to six years, giving machine learning researchers a rich foundation for testing orbit prediction algorithms, anomaly detection, and classification models.
The dataset is explicitly designed to support method development for space domain awareness in the cislunar regime. That distinction matters. Most existing orbital datasets cover low Earth orbit, where physics is simpler and observations are plentiful. By releasing a large, high-fidelity cislunar set under open terms, the authors are trying to accelerate research that would otherwise stall for lack of training data. The practical payoff could be AI systems that predict conjunction risks months or years ahead, giving mission planners enough lead time to adjust trajectories and coordinate among international partners.
Those efforts sit alongside broader attempts to make space science more accessible to data scientists and AI specialists. NASA’s own digital platforms, such as its main online hub and curated series collections, increasingly highlight missions, videos, and related resources that outside teams can build upon. While most of that material focuses on missions closer to Earth, the same open-data philosophy is now beginning to reach the cislunar domain.
Intelligence Agencies Join the Effort
The challenge has also drawn attention from the U.S. intelligence community. The Intelligence Advanced Research Projects Activity issued a request for information on orbital debris detection and tracking, signaling direct interest in advanced computational and AI methods for finding debris that current systems struggle to detect. Separately, IARPA’s SINTRA program seeks to understand how orbital debris interacts with the surrounding space environment, a question with both scientific and intelligence value.
The intelligence angle adds a layer that pure science missions do not typically address. Distinguishing active satellites from debris, and debris from potential threats, requires pattern recognition at a scale and speed that only machine learning can realistically deliver. IARPA’s involvement suggests the U.S. government views cislunar awareness not just as an environmental or safety concern but as a strategic one, particularly as other nations ramp up their own lunar programs and begin to operate in orbits that are hard to monitor with legacy systems.
Earthquake Sensors as Debris Trackers
While most AI efforts focus on space-based or telescope-based detection, a separate line of research tackles the problem from the ground up, literally. Scientists have demonstrated that earthquake sensors can detect falling space junk as it re-enters the atmosphere. The thousands of pieces of human-made debris orbiting Earth pose a direct risk when they fall, and existing seismometer networks could provide a low-cost way to track re-entry events in near real time.
This approach repurposes instruments that are already deployed worldwide, which means the infrastructure cost is minimal. The seismic signatures of debris breaking apart during re-entry are distinct enough for automated systems to identify, opening the door to AI-assisted monitoring that runs continuously without dedicated space hardware. For populated areas beneath common re-entry corridors, that kind of early detection could eventually feed into public safety warnings and post-event analysis of where fragments may have landed.
These ground-based techniques complement satellite and telescope observations rather than replacing them. Just as AI now helps synthesize measurements from Earth science missions with ground networks, combining seismic data with orbital predictions could yield more accurate reconstructions of how debris behaves from orbit to impact. Over time, that knowledge may inform better end-of-life plans for satellites and upper stages, reducing the chance that large pieces survive to the surface.
What Current Coverage Gets Wrong
Public discussion of space junk often focuses on spectacular images of crowded low Earth orbit, but that framing can obscure how different the cislunar problem really is. Near Earth, debris risks are dominated by sheer numbers and high relative speeds. Between Earth and the Moon, the main challenges are sparse observations, complex gravitational dynamics, and the long time horizons over which orbits can drift into conflict. Treating cislunar space as just “a higher orbit” misses the need for new sensing strategies, new datasets, and AI tools tuned to that environment.
Media narratives also tend to separate scientific, commercial, and security concerns into different stories. In reality, those threads are tightly intertwined. The same models that help scientists plan safe trajectories for robotic explorers traveling through the broader universe of deep-space missions can also flag unusual maneuvers by uncooperative objects. Open datasets that enable academic research can simultaneously improve the situational awareness of national space agencies. And techniques first developed to monitor climate and land use through Earth-observing satellites now inform how AI parses faint optical signatures of distant spacecraft.
Another misconception is that debris tracking is purely a technical issue. In practice, policy and coordination will determine how far AI-driven tools can go. Shared catalogs, transparency about planned maneuvers, and agreements on how to handle close approaches will all shape the risk landscape in cislunar space. Technology can highlight impending problems, but only governance can decide who moves, who pays, and how responsibilities are divided among nations and private operators.
For now, the emerging picture is cautiously optimistic. Researchers are building visibility maps to guide telescopes, releasing massive simulated orbit libraries to train machine learning systems, and experimenting with unconventional sensors that hear debris as it falls. Intelligence agencies are signaling that they take the challenge seriously, and open-data initiatives are beginning to bridge communities that once worked in isolation. If those threads continue to converge, the region between Earth and the Moon could become far less of a blind spot just as humanity’s presence there begins to grow.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
