Differential LEO Navigation under Asynchronous Satellite Clocks: Architecture and Performance Bounds

Technical Summary: Differential LEO Navigation under Asynchronous Satellite Clocks

Problem Statement
Low Earth Orbit (LEO) communication satellites offer a promising source of signals of opportunity (SoOP) for positioning, navigation, and timing (PNT), particularly in environments where Global Navigation Satellite Systems (GNSS) are unreliable. However, unlike GNSS constellations, commercial LEO systems lack global time synchronization. Their independent satellite clock biases and drifts introduce severe statistical inconsistencies and degrade kinematic estimation when not explicitly managed.

Existing approaches face significant limitations:

• Single-Receiver Baselines: Treating satellite clocks as nuisance parameters or ignoring them leads to filter inconsistency or weak observability. Explicitly augmenting the rover state with per-satellite clock states inflates the state vector and creates weak observability under time-varying visibility.
• Traditional Differential Methods: Standard differential techniques often assume short baselines where atmospheric effects cancel perfectly and rely on algebraic differencing to eliminate clock errors. These methods fail to account for the rapid temporal dynamics and severe stochastic drift inherent in low-cost, uncooperative LEO oscillators.
• Theoretical Gaps: Existing performance bounds (Cramér-Rao Lower Bounds) are often static, posed at abstract measurement levels, or fail to quantify the impact of injecting a tracked prior (from a base station) on the rover's fundamental error limits.

Methodology
The paper proposes a base-station-aided differential navigation architecture designed to mitigate asynchronous LEO satellite clocks without inflating the mobile rover's state vector. The system operates in two coupled stages:

1. Base Station (BS) Clock-Tracking Filter:

   • A fixed base station with a known position and disciplined time reference continuously tracks per-satellite clock bias and drift states using delay and Doppler measurements.
   • The BS employs a robust visibility management scheme to handle frequent satellite entry and exit. This involves marginalizing lost satellites, permuting states for persistent satellites, and initializing new satellites with a tuned prior covariance to prevent filter overconfidence.
   • The BS outputs per-satellite clock estimates and their rigorously propagated posterior covariance matrices.

2. Rover Kinematic Navigation Filter:

   • The rover utilizes a compact 8-state Extended Kalman Filter (EKF) (3D position, 3D velocity, clock bias, and clock drift). It does not augment its state with satellite clock states.
   • Measurement-Domain Correction: The rover applies the BS's clock estimates directly to its delay and Doppler measurements.
   • Uncertainty Propagation: Crucially, the BS's estimation uncertainty (the posterior covariance of the clock states) is propagated into the rover's measurement covariance matrix. This ensures the rover filter remains statistically consistent and avoids overconfidence, even though it does not explicitly estimate the satellite clocks.

3. Theoretical Performance Bounds:

   • The authors derive the Recursive Bayesian Cramér-Rao Bound (RBCRB) directly linked to the channel-domain signal model.
   • This derivation explicitly quantifies the impact of the tracked prior (the BS's stochastic clock covariance injection) on the rover's information matrix, establishing fundamental theoretical performance limits for dynamic, asynchronous LEO navigation.

Key Contributions

1. Base-Station-Aided Architecture: A differential framework that offloads the tracking of asynchronous LEO clocks to a fixed base station, enabling real-time positioning for dynamic rovers while maintaining a computationally efficient 8-state EKF.
2. Covariance-Aware Correction Protocol: A mathematically rigorous protocol that integrates BS clock estimates into the rover's filter by formally propagating the BS's 2×2 posterior covariance matrix into the rover's measurement update, explicitly mitigating filter overconfidence.
3. Theoretical Performance Benchmarks: The formulation of system-specific analytical Fisher Information Matrices (FIM) to establish RBCRBs that account for the stochastic injection of clock-state covariance from a ground reference station.

Results
The proposed methodology was validated using real-world urban vehicular data (Kingston, Ontario) combined with high-fidelity LEO constellation simulations (XONA PULSAR constellation).

• Statistical Consistency: The differential framework completely eliminates the statistical inconsistency endemic to single-receiver baselines. In the rover-only mode, unmodeled clock errors caused the filter to become overconfident, with errors frequently violating confidence intervals.
• Performance Tracking: The rover's posterior estimation uncertainty in the BS-aided mode perfectly tracks the theoretical RBCRB limits across all kinematic (position, velocity) and clock (bias, drift) states.
• Error Reduction: The BS-aided mode reduced the rover's clock bias error by roughly an order of magnitude compared to the rover-only baseline. Position errors were similarly suppressed and better aligned with predicted uncertainties.
• Robustness: The system maintained stability and accuracy even during periods of severe geometric degradation (high GDOP) and frequent satellite visibility changes, where snapshot bounds would otherwise "blow up."

Significance and Claims
The paper claims that its proposed architecture provides a statistically consistent solution for standalone LEO-based PNT in the absence of global time synchronization. By explicitly managing the stochastic nature of uncooperative LEO clocks through a differential, covariance-aware approach, the system ensures reliable PNT even under challenging geometric conditions. The derivation of the RBCRB linked to the channel model provides a necessary benchmark for evaluating future LEO navigation algorithms, demonstrating that the proposed method operates at the fundamental theoretical limits of information extraction. The work highlights that while differential assistance improves accuracy, the rigorous propagation of uncertainty is essential to maintain filter consistency in dynamic environments.