⚡ electrical engineering

Satellite Autonomous Clock Fault Monitoring with Inter-Satellite Ranges Using Euclidean Distance Matrices

This paper proposes a novel, position-independent framework for detecting satellite clock phase jumps in lunar constellations by analyzing the singular values of geometric-centered Euclidean distance matrices derived from inter-satellite range measurements within rigid graph substructures.

Original authors: Keidai Iiyama, Daniel Neamati, Grace Gao

Published 2026-05-05

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Keidai Iiyama, Daniel Neamati, Grace Gao

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

Imagine a group of satellites orbiting the Moon (or Earth) working together like a team of hikers in a dense forest. To know exactly where they are, they constantly shout distances to one another using radio signals. If everyone is healthy, these distance measurements form a perfect, rigid geometric shape, like a sturdy 3D puzzle.

However, sometimes a satellite's internal clock suddenly "jumps" or skips a beat. This is like one hiker suddenly shouting a distance that is slightly wrong. If the team relies only on comparing their shouts to a map they were given beforehand (the "ephemeris"), they might miss the error if the map itself isn't perfectly accurate.

This paper proposes a new way for the satellites to catch these clock jumps on their own, without needing a perfect map or a ground station to tell them what's wrong. Here is how it works, broken down into simple concepts:

1. The "Rigid Puzzle" Analogy

Think of the satellites as dots and the distance measurements between them as sticks connecting the dots.

The Healthy State: If the sticks are the correct lengths, the dots form a rigid shape that cannot wiggle or bend. In math terms, this is called a "rigid graph."
The Broken State: If one satellite's clock jumps, it lies about the length of the sticks connected to it. Suddenly, the puzzle pieces no longer fit together. The shape becomes "unrealizable"—it's like trying to build a square table where one leg is suddenly 10 inches longer than the others; the table just won't stand up.

The paper argues that if you have a specific, tightly connected group of five satellites (a "5-clique"), you can mathematically prove that if their distances don't fit together perfectly, someone in that group must be lying about the distance.

2. The "Magic Mirror" (Euclidean Distance Matrices)

How do the satellites know the puzzle is broken without solving for every single position? They use a mathematical tool called a Geometric-Centered Euclidean Distance Matrix (GCEDM).

Imagine this matrix as a "magic mirror" that looks at all the distance measurements at once.

When everything is normal: The mirror shows a very specific, clean pattern (mathematically, it has a low "rank," meaning it fits neatly into 3D space).
When a clock jumps: The lie introduced by the bad clock distorts the pattern. The mirror suddenly shows "noise" or extra dimensions that shouldn't be there.

The authors developed a test that looks at the fourth number in a list of values (singular values) generated by this mirror.

If the number is tiny (close to zero), the puzzle is fine.
If the number is big, the puzzle is broken, and a clock jump has occurred.

3. Dealing with Sparse Connections

In space, satellites can't always see each other (they might be blocked by the Moon or Earth). This means they can't always form that perfect "5-satellite puzzle."

To fix this, the paper suggests a hybrid approach:

If the satellites can't measure a distance directly, they can use a "best guess" based on their known orbital paths (ephemeris) to fill in the missing stick in the puzzle.
However, they only use these "guesses" to connect the dots. The actual "lie detector" test still relies on the real, measured distances. As long as every satellite has at least one real connection to the group, the system can still spot the liar.

4. The Results

The authors tested this idea using simulations of:

Earth's GPS system: A dense network with many satellites.
A notional Moon network: A smaller, sparser network.

They found that their method works very well:

It can detect clock jumps even when the network is sparse (few connections).
It is often better than older methods that rely heavily on the accuracy of the orbital maps.
It can tell you exactly which satellite is the "liar" by checking which one, when removed from the group, makes the puzzle fit again.

Summary

In short, this paper teaches a fleet of satellites to act like a group of friends checking each other's work. Instead of trusting a potentially flawed map, they use the geometry of their own connections. If the distances they measure don't add up to a solid 3D shape, they know a clock has jumped, and they can pinpoint exactly who made the mistake, even if they can't see everyone all the time.

Technical Summary: Satellite Autonomous Clock Fault Monitoring with Inter-Satellite Ranges Using Euclidean Distance Matrices

Problem Statement
Reliable Positioning, Navigation, and Timing (PNT) services for lunar constellations (e.g., LunaNet) require robust monitoring of satellite clock health. Traditional integrity monitoring methods face limitations in the lunar environment: ground-based monitoring is impractical due to the lack of surface infrastructure; Receiver Autonomous Integrity Monitoring (RAIM) requires a minimum of five to six visible satellites, which may not be available in smaller lunar constellations; and existing Satellite Autonomous Integrity Monitoring (SAIM) methods often rely on comparing Inter-Satellite Range (ISR) measurements against ephemeris predictions. The latter approach's fault detection sensitivity is heavily dependent on ephemeris accuracy, which is challenging to maintain for lunar orbits. Furthermore, many current SAIM methods process measurements on a per-satellite basis, failing to fully exploit the geometric redundancy of the entire constellation.

Methodology
The authors propose a novel framework for detecting satellite clock phase jumps using dual one-way ISR measurements, grounded in rigid graph theory and Euclidean Distance Matrix (EDM) analysis.

Measurement Model: The method utilizes dual one-way ISR measurements between satellite pairs. By combining forward and reverse links, the authors derive "clock-free" and "geometry-free" observation equations. A clock phase jump on a satellite introduces a structured bias into these range measurements.
Graph-Theoretic Foundation: The constellation is modeled as a weighted graph where nodes represent satellites and edges represent measured ranges. The core hypothesis is that a set of range measurements is "realizable" (embeddable) in 3D Euclidean space only if the underlying graph is rigid. Specifically, the authors establish that vertex-redundant rigidity is required for unambiguous fault detection. A graph is $k$ -vertex redundantly rigid if it remains rigid after removing any $k-1$ vertices. For 3D space, a fully connected graph of 5 nodes (a 5-clique) is the minimal structure that is 2-vertex redundantly rigid.
EDM and GCEDM Analysis: Instead of solving for satellite positions, the method constructs a Geometric-Centered Euclidean Distance Matrix (GCEDM) from the observed ranges.
- Theoretical Proof: The authors prove that for a noiseless, fault-free graph in 3D, the rank of the GCEDM is at most 3. If a clock fault exists, the rank increases.
- Statistical Test: In the presence of measurement noise, the 4th singular value of the GCEDM becomes non-zero. The paper derives the analytical distribution of the squared 4th singular value ( $\lambda_4^2$ ) under both nominal (noise-only) and faulty conditions. Under nominal conditions, the test statistic follows a central Chi-squared distribution; under faulty conditions, it follows a non-central Chi-squared distribution.
Algorithm:
- Clique Listing: The algorithm identifies all 5-clique subgraphs within the constellation topology.
- Hypothesis Testing: For each 5-clique, the squared 4th singular value is computed and compared against a threshold derived from the noise variance.
- Fault Identification: To identify the specific faulty satellite, the algorithm sums the test statistics of all subgraphs excluding a specific satellite. The satellite whose exclusion results in the lowest aggregate test statistic is identified as the fault source.
Ephemeris Augmentation: To address sparse connectivity (common in lunar constellations where 5-cliques may not exist naturally), the authors introduce an augmentation method. Missing edges in the graph are "filled in" using range estimates derived from satellite ephemerides. This creates hybrid subgraphs (measured + computed edges) that satisfy the fault-detectable topology, provided every satellite has at least one measured edge incident to it.

Key Contributions

Graph-Topological Conditions: The paper establishes the necessary condition of vertex-redundant rigidity (specifically 5-cliques in 3D) for unambiguous fault detection from ISR measurements without prior knowledge of satellite positions.
Statistical Derivation: The authors provide rigorous proofs regarding the rank properties of EDMs and GCEDMs in the presence of faults and noise. They derive the analytical distribution of the test statistic (squared 4th singular value) for $n=5$ , characterizing both the scale (dependent on noise) and the non-centrality parameter (dependent on fault magnitude).
Novel Algorithm: A hypothesis testing framework is developed that detects and identifies clock phase jumps by monitoring the realizability of subgraphs via GCEDM singular values, independent of ephemeris accuracy for the core detection logic.
Hybrid Formulation: An ephemeris-augmented approach is proposed to extend fault detection capabilities to sparse constellations by combining measured and computed ranges, trading off some sensitivity to ephemeris errors for increased availability.

Simulation Results
The method was validated using Monte Carlo simulations on two scenarios: a 31-satellite GPS constellation and a notional 9-satellite lunar constellation.

Performance Comparison: The proposed EDM-based method was compared against two baselines: "ephemeris comparison" (residual-based) and "Data Snooping" (Baarda's w-test).
GPS Constellation: The EDM method demonstrated robust performance, with a fault detection capability (Minimum Detectable Bias) intermediate between the optimal Data Snooping method and the ephemeris comparison method. It showed lower false alarm rates than the ephemeris method when ephemeris accuracy was poor.
Lunar Constellation: In the sparse lunar scenario, the ephemeris-augmented EDM method successfully detected faults where pure ISR-based methods would fail due to a lack of 5-cliques. The results indicated that while the method's sensitivity depends on ephemeris accuracy in the augmented mode, it remains effective across a wide range of fault magnitudes and link topologies.
Sparse Faults: When only a subset of links connected to a faulty satellite were biased (fault ratio = 0.2), the EDM method maintained better robustness than Data Snooping, which relies on the assumption that all links to a faulty satellite are corrupted.

Significance and Claims
The paper claims to offer a new theoretical and practical perspective on fault detection by viewing the problem through the lens of graph rigidity and geometric realizability rather than simple residual analysis. The authors state that their method is not intended to replace decades of existing research but to complement it. The primary significance lies in the ability to perform autonomous fault detection that is less dependent on high-precision ephemeris data (in the pure ISR mode) and capable of operating in sparse lunar constellations (via the augmented mode). The work provides a statistically rigorous framework for identifying clock phase jumps using inter-satellite links, addressing a critical gap in the development of reliable lunar PNT infrastructure.

1. The "Rigid Puzzle" Analogy

2. The "Magic Mirror" (Euclidean Distance Matrices)

3. Dealing with Sparse Connections

4. The Results

Summary

More like this