Reduced-Dynamic Filtering for GNSS-Only Orbit Determination in Cislunar Space: An Experimental Assessment with LuGRE Data

Global Navigation Satellite Systems (GNSSs) are increasingly regarded as an enabling technology for onboard Positioning, Navigation and Timing (PNT) beyond the Space Service Volume (SSV), where navigation autonomy is becoming central to sustained lunar exploration. In cislunar space, however, GNSS-based navigation is challenged by intermittent satellite visibility and poor measurement geometry, motivating the development of Orbit Determination and Time Synchronization (ODTS) architectures that remain reliable under sparse radiometric conditions while being computationally feasible for onboard use. Reduced-dynamic filtering provides a practical compromise by combining a physics-based orbit propagator with stochastic dynamic compensation, offering robustness against dynamic mismodeling compared to purely-dynamic approaches. This work presents the first experimental assessment of GNSS-only reduced-dynamic ODTS in cislunar space using real spaceborne data from the Lunar GNSS Receiver Experiment (LuGRE). Raw pseudorange and Doppler-shift measurements, recorded by the GNSS payload during three representative operations across the Earth-Moon phasing orbits and on the lunar surface, are processed using a first-order Extended Kalman Filter (EKF). To account for the different perturbation environments in each orbit regime, the ODTS filter incorporates a distance-adaptive lunar gravity model, ensuring physical consistency across Earth-Moon regimes. The results obtained across three LuGRE operations, spanning geocentric distances from 15.03 to 51.69 Earth radii (RE) and including one surface operation, demonstrate the viability of pseudorange-based reduced-dynamic filtering despite poor satellite geometry and depleted observability. In the most favorable low altitude scenario, the position error reaches 1.36 km at the 95th percentile. In the more challenging cislunar and lunar-surface operations, the orbit estimation accuracy remains within 3.91 km. Velocity errors are below 1 m/s across all cases. While the filter provides stable and continuos solutions even under degraded conditions, the analysis reveals sensitivity to residual systematic measurement errors, which can induce biased and inconsistent estimates. These preliminary results provide an experimentally grounded benchmark for GNSS-based ODTS in cislunar space and motivate further improvements in measurement modeling and adaptive process noise tuning.