⚡ electrical engineering

System Level Analysis and Management of Orbital Debris Using Empirical Dynamic Modeling

This paper proposes a data-driven approach using Empirical Dynamic Modeling to reconstruct the orbital debris system's shadow attractor from limited time-series data, enabling the simulation of future debris scenarios and the assessment of policy impacts on space operations.

Original authors: Asaad S. Abdul-Hamid, Hao Chen

Published 2026-05-22

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Asaad S. Abdul-Hamid, Hao Chen

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: Space is Getting Crowded

Imagine Earth's orbit as a busy highway around our planet. Over the last 60 years, we've added thousands of cars (satellites), but we've also left behind a lot of trash: broken parts, old fuel tanks, and pieces from crashes. This "space junk" is dangerous. If a piece of trash hits a working satellite, it can destroy it and create even more trash, leading to a chain reaction of collisions.

Scientists and policymakers want to know: How bad will this get? And if we change the rules (like making satellites leave orbit faster), will it help?

The Problem with Old Tools

Usually, to predict the future of this space highway, experts build complex mathematical models. They try to write down every single rule of physics and human behavior.

The Analogy: Imagine trying to predict the weather by writing a giant equation for every single water molecule in the atmosphere. It's incredibly hard, requires perfect data, and if you miss one rule, your prediction fails.
The Paper's Issue: We don't have perfect data on space junk. We only have a few data points (about 60 years of yearly counts). Traditional models struggle with this lack of information and the chaotic nature of space.

The New Solution: "Empirical Dynamic Modeling" (EDM)

The authors propose a new way to look at the problem. Instead of guessing the rules, they look at the pattern of the data itself.

The Analogy: The Shadow on the Wall
Imagine a complex 3D sculpture (the real space system) spinning in a dark room. You can't see the whole sculpture, but you have a light shining on it, casting a shadow on the wall.

The paper uses a mathematical trick (called Takens' Embedding Theorem) that says: If you study the shape of the shadow closely enough, you can figure out what the 3D sculpture looks like, even without seeing the sculpture itself.
By looking at the "shadow" of the data (how the number of debris objects changes year by year), they can reconstruct the hidden "shape" or attractor of the whole system. This allows them to see how the system moves without needing to know every single rule of physics.

How They Tested It

The researchers used three main tools, which they compare to different ways of reading a story:

Simplex Projection (The "Look Back" Method):
- They looked at past years of data and asked, "What happened in the past that looks most like today?" Then they assumed the future would follow that same path.
- Result: This worked okay for matching past data, but it failed to predict the future. It was like looking at a flat line and assuming the future would be flat. It missed the fact that space junk tends to multiply itself (like a snowball rolling downhill).
Convergent Cross Mapping (The "Detective" Method):
- They wanted to know: Does the number of new satellites launched actually cause more debris? Or is it just a coincidence?
- They used a test to see if the "story" of the launches and the "story" of the debris were written by the same author.
- Result: They found a strong, two-way connection between the total number of objects in space and the amount of debris. They confirmed these variables are causally linked.
S-Mapping (The "Smart Predictor"):
- This is the upgraded version of the first method. Instead of just looking at the closest neighbor in the past, it looks at the whole history of the data, weighing the most relevant parts more heavily.
- Result: This method successfully predicted that the debris population will grow significantly, reaching over 30,000 pieces by 2050. It was much better at capturing the "non-linear" (chaotic) nature of the problem than the simple method.

Testing Policy Changes (The "What If" Scenarios)

Once they built a reliable model, they simulated different policies to see which ones would work best. They imagined these policies starting in the year 2000 and looked at the results in 2050.

The "Post-Mission Disposal" (PMD) Policy:
- The Idea: Satellites must leave orbit faster after they stop working (e.g., within 5 years instead of 25).
- The Result: This was the most effective strategy. Shortening the time a dead satellite stays in space drastically reduced the predicted debris. Interestingly, a 10-year limit worked better than a 5-year limit in their simulation, likely because the timing of the change mattered more than the intensity.
The "Launch Less" Policy:
- The Idea: Reduce the number of new satellites launched by 10% to 40%.
- The Result: This helped, but not as much as the PMD policy. The paper notes that because the space industry is booming (with "mega-constellations"), simply launching fewer satellites is the least realistic option.
The "Active Debris Removal" (ADR) Policy:
- The Idea: Send robots to physically catch and remove 100 to 3,000 pieces of trash every year.
- The Result: This had a very small impact. Even removing 3,000 pieces a year only reduced the total debris by about 16% compared to doing nothing. The paper suggests this is because the model treats all debris as equal, whereas in reality, removing specific "dangerous" pieces would be more effective. Also, the model showed that if you stop removing trash, the population quickly returns to its original growth path.

The Bottom Line

The Method: The paper successfully used a data-driven "shadow" method to understand the complex, chaotic system of space debris without needing perfect physics equations.
The Prediction: Without major changes, space debris will continue to grow rapidly.
The Best Solution: The most effective way to stop the growth is to make sure dead satellites leave orbit quickly (Post-Mission Disposal).
The Limitation: The model can only predict based on patterns seen in the past. If something totally new happens (like a massive, sudden explosion or a completely new type of satellite behavior) that hasn't happened in the historical data yet, the model might not catch it immediately.

In short, the authors built a new "crystal ball" based on data patterns rather than physics equations, and it tells us that cleaning up space by making satellites leave orbit faster is our best bet.

Technical Summary: System Level Analysis and Management of Orbital Debris Using Empirical Dynamic Modeling

Problem Statement
Orbital debris presents a critical threat to global space operations and the sustainability of the space economy. As of 2022, over 15,000 debris objects in Low Earth Orbit (LEO) pose collision risks to active satellites, with the potential for catastrophic fragmentation events (Kessler syndrome) to exponentially increase debris populations. While existing literature addresses debris through policy, technical remediation, and economic cost-benefit analyses, these approaches often rely on parametric models requiring extensive assumptions about system mechanics or extensive historical data. Furthermore, traditional models struggle to capture the complex, nonlinear, and bidirectional dynamics of the debris environment, particularly when data is limited (e.g., only ~60 annual data points since 1960) and system states are non-equilibrium. There is a distinct need for tools that can understand system-level implications of debris solutions using limited observable variables without prior knowledge of the underlying physical equations.

Methodology
The authors propose a data-driven framework based on Empirical Dynamic Modeling (EDM), rooted in Takens' Embedding Theorem. This approach treats the orbital debris environment as a dynamic system where variables (debris count, total objects, launch rates) share a common system attractor manifold. The methodology proceeds through four key stages:

Attractor Reconstruction: Instead of hypothesizing parametric measurement functions ( $\Phi$ ) or dynamic evolution functions ( $\Psi$ ), the method uses time-series monitoring data to reconstruct a "shadow attractor." By utilizing delay coordinate mapping with an embedding dimension ( $E$ ) and time delay ( $\tau$ ), the system's state space is recovered from observable variables (e.g., annual debris counts).
Simplex Projection: This technique forecasts future states by identifying nearest neighbors on the reconstructed shadow manifold and projecting them forward. It is used initially to determine the optimal embedding dimension ( $E$ ) and assess prediction skill ( $\rho$ ).
Convergent Cross Mapping (CCM): This method verifies causal relationships between system variables. By testing if the state space reconstruction of one variable (e.g., total objects) can predict another (e.g., debris count), the authors confirm bidirectional causality, ensuring the variables belong to the same dynamic system.
S-Mapping (Sequential Locally Weighted Global Linear Mapping): Recognizing that Simplex Projection (which relies only on nearest neighbors) failed to capture the self-propagating nature of debris collisions, the authors employ S-Mapping. This method performs locally weighted linear regression across the entire dataset, allowing for the analysis of nonlinearity via a localization parameter ( $\theta$ ). S-Mapping provides partial derivatives to quantify the strength and sign of interactions between variables.

Key Results

System Dynamics: CCM analysis confirmed a strong bidirectional causal link between the number of debris objects in LEO and the total number of objects in LEO. The optimal embedding dimension was determined to be $E=4$ .
Forecasting Performance: While Simplex Projection ( $E=4$ ) achieved a 96.6% prediction skill, it unrealistically predicted a flattening debris population. S-Mapping, with an optimal localization parameter of $\theta=7$ , achieved a 97.1% prediction skill and correctly forecasted a continued increase in debris, predicting over 30,000 debris objects (>10 cm) in LEO by 2050 under current trends. This data-driven forecast exceeded the mean of a 25,000-run Monte Carlo simulation using the DAMAGE evolutionary model, suggesting the inclusion of complex dynamics like explosions and collisions not fully captured by standard parametric models.
Policy Simulations: The authors simulated three mitigation scenarios effective from the year 2000:
- Post-Mission Disposal (PMD): Reducing the post-mission disposal lifetime (e.g., from 25 years to 5 or 0 years). The 10-year PMD policy showed the highest effectiveness (51.14% mitigation by 2050), outperforming the 5-year policy due to the specific cadence of deorbiting events interacting with the system dynamics.
- Launch Reduction: Reducing annual launches by up to 40%. This yielded a maximum of 19.55% mitigation but was noted as less realistic given the growth of mega-constellations.
- Active Debris Remediation (ADR): Removing 100 to 3,000 random debris pieces annually. Even at 3,000 removals/year, mitigation was only 16.59%. The simulation showed that without targeting specific high-risk objects, ADR has diminishing returns as the population rebounds.

Significance and Claims
The paper claims to represent a significant advancement in the ability to understand high-level impacts of space system policy using limited data. By framing orbital debris as a dynamic process rather than a static inventory, the proposed EDM framework enables:

Causal Discovery: Identifying fundamental system dynamics and causal links without requiring prior mechanical understanding of the system.
Policy Quantification: Providing a quantitative basis for how specific policy changes (like PMD rules) alter the future trajectory of the debris population.
Data Efficiency: Successfully reconstructing system mechanics and forecasting future states with only ~60 data points, a feat difficult for traditional parametric models.

The authors conclude that while the method has limitations—specifically its reliance on past empirical data patterns and inability to weigh specific debris objects by risk—it offers a novel, equation-free tool for holistic system analysis. The results suggest that reducing post-mission disposal lifetimes is currently the most viable and effective mitigation strategy compared to launch reductions or random active debris removal. Future work is suggested to incorporate more diverse time-series variables (e.g., mass of launches, specific debris events) and to analyze economic and environmental impacts of debris solutions.