🔢 mathematics

Geometric Analysis of Doppler-Based Navigation with Low Earth Orbit Satellites

This paper establishes a geometric foundation for Doppler-based navigation using Low Earth Orbit satellites by deriving a closed-form parameterization of the measurement Jacobian and proving that while satellite altitude diversity reduces clock bias-coupling, the inherent geometric correlation between clock bias and clock drift fundamentally inflates the geometric dilution of precision, rendering traditional volume-based satellite selection ineffective.

Original authors: Carlos Caravaca Gallego, Pini Gurfil, Hector Rotstein

Published 2026-09-11

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Carlos Caravaca Gallego, Pini Gurfil, Hector Rotstein

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 trying to find your way using only the changing pitch of a passing siren. If you know how fast the siren is moving and how the sound shifts as it approaches and recedes, you can calculate exactly where you are and how fast you are moving. This is the principle behind using Low Earth Orbit satellites for navigation. Unlike the familiar Global Navigation Satellite Systems that rely on measuring the time it takes for a signal to travel, this method listens to the Doppler shift—the change in frequency of the radio signal caused by the satellite's rapid motion. Because these satellites circle the Earth much closer than traditional navigation satellites, their signals are significantly stronger, offering a potential lifeline if the standard systems are jammed or fail. However, turning these frequency shifts into a precise location is a complex puzzle involving eight different unknowns: where you are, how fast you are moving, and the tiny errors in your receiver's clock.

For years, engineers have relied on a simple rule of thumb to pick the best satellites for this job: choose the ones that form the widest possible shape in the sky. In traditional navigation, spreading satellites out maximizes the volume of space they cover, which usually leads to the most accurate position. But when researchers applied this same logic to Doppler-based navigation, it failed. The satellites that looked best for traditional systems often produced poor results for Doppler navigation. This disconnect puzzled experts until a new study provided the answer, revealing that the geometry of Doppler navigation is fundamentally different because of how the receiver's clock error interacts with the satellite's motion.

The researchers, working at the Technion in Israel, set out to understand why the old rules didn't work and to find a new way to select the best satellites. They developed a detailed mathematical map of how the Doppler signal changes based on the satellite's height, its direction of travel, and its position in the sky relative to the observer. Their analysis showed that the error in the receiver's clock is not independent of the satellite's position, as it is in traditional systems. Instead, the clock error is tightly linked to the satellite's speed and its angle in the sky. Specifically, the way the clock error affects the measurement depends heavily on how high the satellite is and the angle between its path and the line of sight to the receiver. This creates a hidden connection, or correlation, between the clock error and the other variables the system is trying to solve for.

This connection is the reason the old "volume" method fails. In traditional navigation, the clock error column in the math is a straight line that doesn't change with the satellite's position. In Doppler navigation, that line bends and shifts depending on where the satellite is. The researchers proved that this bending creates an unavoidable overlap between the clock error and the other measurements. No matter how you arrange the satellites, you cannot completely separate the clock error from the position and velocity data. This overlap inflates the uncertainty of the final result, making the position less precise than it would be if the clock error could be isolated perfectly.

To solve this, the team looked for a way to break this link. They discovered that the most effective way to reduce the overlap is to use satellites at different altitudes. When satellites are all at the same height, they tend to produce similar patterns of error, reinforcing the correlation. But if you mix satellites from different orbital heights—some closer to Earth and some farther away—they create distinct patterns of sensitivity. This difference in height helps separate the clock error from the other variables, reducing the uncertainty. However, there is a catch. Satellites closer to the Earth provide much stronger signals and more information per measurement, while those farther away provide weaker signals. The optimal solution, the researchers found, is not to spread satellites evenly across all heights, but to use a "bang-bang" approach: place most satellites at the lowest possible altitude to maximize signal strength, and use just a few satellites at a much higher altitude to break the correlation.

In a simulation using a hypothetical constellation of eight satellites, the researchers tested this theory. They found that a configuration with seven satellites at a low altitude of 400 kilometers and one satellite at a high altitude of 1,200 kilometers produced the best results. This mix reduced the uncertainty significantly more than using all satellites at the same height. The study confirmed that while the old volume-based selection method is useless for Doppler navigation, a new strategy based on altitude diversity can work. The findings suggest that future navigation systems relying on Doppler shifts will need to carefully mix satellites from different orbital layers to achieve the highest precision, rather than simply picking the ones that look best from the ground. The research provides a clear geometric foundation for this selection, showing that the key to accuracy lies in understanding how the satellite's height changes the way its motion reveals the receiver's clock error.

Technical Summary: Geometric Analysis of Doppler-Based Navigation with Low Earth Orbit Satellites

Problem Statement
The increasing vulnerability of Global Navigation Satellite Systems (GNSS) has renewed interest in Low Earth Orbit (LEO) satellite navigation. While large LEO constellations (e.g., Starlink, OneWeb) offer strong signal availability, tracking all visible satellites is computationally expensive, necessitating efficient satellite selection. In pseudorange-based navigation, Geometric Dilution of Precision (GDOP) minimization is effectively approximated by maximizing the volume spanned by line-of-sight (LOS) unit vectors. However, empirical studies (e.g., Moore et al., 2024) indicate that this volume-based heuristic fails for Doppler-based navigation, where the goal is to estimate eight states (three position, three velocity, clock bias, and clock drift) solely from carrier Doppler measurements. The failure of volume-based selection suggests a fundamental structural difference in the Doppler measurement Jacobian, specifically regarding the clock bias column, which has not been analytically characterized in closed form.

Methodology
The paper establishes a geometric foundation for Doppler GDOP by deriving a closed-form parameterization of the eight-state Doppler measurement Jacobian. The authors express the Jacobian elements in terms of topocentric variables: elevation ( $\varepsilon$ ), azimuth ( $\alpha$ ), orbital inclination ( $i$ ), and altitude ratio ( $\beta = R_E/a_{orb}$ ).

Key methodological steps include:

Closed-Form Derivation: The authors derive exact analytical expressions for the clock bias sensitivity ( $c_j$ ) as a function of elevation, altitude, and the aspect angle ( $\chi$ ) between the satellite velocity vector and the LOS.
Structural Decomposition: The eight-state information matrix is partitioned into a seven-state subspace (position, velocity, clock drift) and the clock bias column.
Schur Complement Analysis: A Schur complement decomposition is applied to the information matrix to derive an exact formula for GDOP inflation. This isolates the cost of estimating clock bias in terms of a collinearity coefficient ( $\zeta^2$ ), which measures the alignment of the clock bias column with the seven-state subspace.
Numerical Validation: A synthetic two-shell constellation (satellites at 400 km and 1200 km) is optimized using particle swarm optimization to verify the theoretical inflation formula and analyze the trade-offs between altitude diversity and sensitivity.

Key Contributions and Results

Geometric Parameterization of Clock Bias Sensitivity: The paper proves that the clock bias sensitivity ( $c_j$ ) is strictly negative for all satellites above the horizon. It is bounded by two envelopes dependent only on elevation and altitude: an in-plane floor and a cross-track ceiling. Crucially, $c_j$ depends on the LOS rate magnitude (shared with position sensitivity) and the elevation factor (shared with the vertical velocity component).
Irreducible Correlation: The analysis demonstrates that the clock bias column is irreducibly correlated with the clock drift column (due to the sign-definiteness of $c_j$ ) and structurally coupled to the position and vertical velocity columns. This coupling is inherent to the physics of Doppler measurements and cannot be eliminated by any geometric arrangement.
Exact GDOP Inflation Formula: The authors derive a closed-form expression for the increase in GDOP when adding the clock bias state:
$GDOP_8^2 - GDOP_7^2 = \frac{\|\xi\|^2 + 1}{d(1 - \zeta^2)}$
where $d$ is the total clock bias sensitivity, $\xi$ represents the projection of the clock bias column onto the seven-state subspace, and $\zeta^2$ is the collinearity coefficient. This formula proves that perfect decorrelation ( $\zeta^2 = 0$ ) is impossible for any constellation with satellites above the horizon.
Altitude Diversity as a Decorrelation Mechanism: The paper identifies altitude diversity as the primary mechanism for reducing collinearity. Satellites at different altitudes but the same elevation possess different clock bias sensitivities, displacing them from the "single-shell band" that forces high collinearity. However, this benefit is limited by a sensitivity-decorrelation tradeoff: higher-altitude satellites offer better decorrelation but weaker Doppler signatures (lower $d$ ), while lower-altitude satellites provide high sensitivity but force high collinearity.
Optimization Results: Numerical optimization of a two-shell constellation reveals a "bang-bang" tendency: optimal configurations place satellites at the extreme altitude bounds (e.g., 400 km and 1200 km) rather than intermediate altitudes. The results confirm that while altitude diversity improves GDOP, the majority of the gain comes from lowering the constellation altitude (increasing $d$ ), with altitude diversity providing a smaller, secondary improvement.

Significance and Claims
The paper claims to provide the missing analytical link explaining why volume-based satellite selection fails for Doppler navigation. By characterizing the clock bias column explicitly, the authors show that the failure arises because the clock bias column is not independent of geometry (unlike in pseudorange navigation) but is instead structurally coupled to position and velocity states.

The significance of the work lies in:

Structural Insight: Establishing that the geometric coupling making clock bias observable is the same coupling that inflates estimation covariance.
Limiting Heuristics: Demonstrating that simple geometric proxies (like LOS volume) are insufficient for Doppler GDOP minimization because they ignore the specific coupling of the clock bias column.
Design Implications: Providing a theoretical basis for satellite selection that prioritizes the full Doppler Jacobian computation over simplified proxies, while highlighting the limited but real benefits of altitude diversity in multi-constellation environments.

The authors maintain a modest scope, focusing on the geometric mechanism governing relative constellation quality rather than absolute error budgets, and note that current LEO altitude separations offer only partial relief from the inherent collinearity.

More like this