- The paper introduces quasi-direct geolocation (QDG) as a complexity-scalable technique that pre-processes I/Q data using STFT, achieving over 99.9% compression while preserving key signal features.
- The paper demonstrates successful localization of a 50W ground-based jammer within a 0.5 km² area using single-satellite multi-antenna experiments, consistent with CRLB predictions.
- The paper validates QDG's feasibility for real-time on-orbit interference mitigation, paving the way for scalable, distributed RF monitoring and enhanced GNSS resilience.
Complexity-Scalable Quasi-Direct Geolocation for Terrestrial GNSS Jammer Detection from LEO
Introduction
The proliferation of terrestrial radio-frequency (RF) interference, including jammers, poses significant threats to Global Navigation Satellite Systems (GNSS) by degrading positioning, navigation, and timing (PNT) reliability. Detecting and geolocating such RF interference sources from low Earth orbit (LEO) satellites remains challenging due to the severe constraints on onboard computational capacity, downlink bandwidth, and the size, weight, and power (SWaP) of most LEO platforms. This work presents a detailed theoretical framework and experimental validation of quasi-direct geolocation (QDG) as a complexity-scalable method for real-time identification and mitigation of terrestrial GNSS jammers using SWaP-constrained LEO satellites, exemplified via the OPS-SAT PRETTY CubeSat during the Jammertest 2025 campaign.
Background and State-of-the-Art
Traditional approaches to RF emitter geolocation—namely, two-step techniques built upon extracting and aggregating physical observables such as time/frequency/angle of arrival and their inter-receiver differences—suffer from fundamental sub-optimality. The reduction of raw I/Q data to intermediate observables discards critical signal information, resulting in degraded sensitivity and ambiguity when multiple, potentially unknown, emitters coexist within the surveillance footprint. Direct geolocation (DG), or direct position determination (DPD), offers improved accuracy by leveraging maximum-likelihood searches in the position domain, directly exploiting the information content in the I/Q samples. However, DG's computational demands, driven by exhaustive grid-searches over large area and emission parameter spaces, render it infeasible for in-situ LEO execution under SWaP limitations, and impractical due to massive downlink requirements.
Prior demonstrations of DG for RFI detection from dual-satellite constellations highlight its superior performance, but also emphasize platform limitations that curtail routine, timely deployment. The QDG paradigm introduced in this paper seeks to bridge the efficiency-accuracy gap: by pre-processing I/Q data into compressed, quantized time-frequency representations (via short-time Fourier and S transforms) and by approximating exhaustive position-domain search methods, QDG enables single- or multi-antenna, single-satellite, low-latency geolocation of major classes of interference signals, most notably narrowband continuous wave (CW) and frequency-swept jammers, with constrained computational overhead.
QDG Mathematical Framework and Algorithmic Structure
QDG operates by pre-processing raw I/Q satellite-collected data into a time-frequency domain through discrete linear transforms such as STFT. Compression is aggressively realized by thresholding time-frequency bins against estimated noise floors, resulting in >99.9% compression for CW-dominated scenarios, while maintaining the signal integrity required for geolocation tasks.
The central cross-correlation (or autocorrelation) operation is approximated using the compressed STFT products between multiple antennas (or over time for a single antenna), parametrized over a geographic grid. The search is accelerated by avoiding computationally expensive fractional delay and Doppler mixing steps, usually required in full DG/DPD, and leveraging the structure of the STFT as a lookup table. When employing a multi-antenna receiver, spatial diversity is exploited through approximate phase difference of arrival (PDOA) measurements, which, after calibration of constant array-specific offsets, enhance directionality and SNR of the geolocation process. For single-antenna scenarios, temporal diversity is used by integrating across acquisition intervals, though at a precision trade-off.
QDG incorporates formal estimation-theoretic bounds, notably the Cramer-Rao lower bound (CRLB), to characterize achievable geolocation precision given platform navigation state errors, signal SNR, receiver noise, geometry, and ancillary constraints (e.g., altitude prior from global digital elevation models). Nominally, horizontal error ellipses on the order of 900 m (major semi-axis, 95% percentile) have been observed in experimental validation.
Signal Isolation, Cancellation, and Multi-Source Disambiguation
A recurring impediment in practical geolocation is the presence of multiple emission sources of differing power, potentially with overlapping spectral signatures. QDG proposes two robust methods—beam/null steering using multi-antenna arrays and time-frequency domain filtering using STFT masks—to sequentially cancel out dominant sources (i.e., high-SNR local jammers), enabling successive isolation and geolocation of weaker, distant emitters that would otherwise be masked. Null steering, though practically limited by the number of antenna elements, enables coherent suppression of signal contributions from specific directions when array geometry is sufficiently well known and calibrated. Time-frequency filtering, in contrast, requires no array and can process an arbitrary number of jamming signals, making it well-suited for CubeSats and single-antenna platforms.
Upon removal of the most powerful jammer in the experimental data set, previously obscured, weaker RF energy contributions attributed to persistent RFI hotspots—such as those affecting the Baltic region from >2000 km range—were rendered localizable, cross-validated against independently collected GNSS integrity metrics.
Experimental Validation: OPS-SAT PRETTY in Jammertest 2025
During coordinated operations at the Jammertest 2025 open-air RFI event, OPS-SAT PRETTY—a 3U CubeSat originally designed for GNSS reflectometry and repurposed for RFI detection—acquired multi-antenna, high-rate I/Q samples during repeated overpasses of a known high-power CW jammer. The on-board SDR with two patch antennas and a precise, though uncalibrated, baseline, generated data subsequently processed with QDG.
Hardware and environmental uncertainties, such as LO frequency offset, antenna phase calibration, and platform navigation state errors, were systematically characterized and mitigated in QDG post-processing. After re-quantization to 8 bits and STFT denoising with an aggressive false-alarm threshold, a 1 GB raw I/Q dataset was reduced to <1 MB, enabling feasible real-time processing or rapid ground downlink.
Key findings include:
- Successful localization of a 50 W ground-based jammer to within a 0.5 km² area, with SNR and error ellipses consistent with theoretical bounds set by the derived CRLB.
- Demonstration of sequential cancellation/reduction of dominant signals and exposure of multiple, geographically distant, low-SNR jammer signals, agreeing with independent GNSS anomaly traces.
- QDG's run-time and data volume reduction facilitate in-orbit execution on embedded CPUs or minimal downlink scenarios, demonstrating the possibilities for SWaP-constrained platform participation in global RFI surveillance.
Implications and Future Directions
The operationalization of QDG for GNSS RFI geolocation yields significant practical implications for both satellite platform design and global GNSS resilience policy. By dramatically reducing data storage and transfer requirements, and by exploiting only modest onboard processing resources, QDG allows the integration of RFI geolocation capabilities into a vast number of new and existing LEO assets, including opportunistically repurposed satellites and future non-terrestrial network (NTN) constellations. Theoretical and experimental results suggest that near real-time global RF interference monitoring and rapid response can be feasibly distributed across orbiting infrastructure, even without persistent on-board GPU resources or dedicated high-speed downlinks.
For future research, multi-satellite QDG fusion, hardware-accelerated onboard implementations, adaptation to wideband and non-CW interferers, and tighter integration with complementary performance-proxy data sources represent immediate technical frontiers. Additionally, extending QDG to real-time, autonomous in-space operation could multiply the density and latency of global jamming detection networks without substantial increments in mission cost or resource overhead.
Conclusion
QDG introduces an efficient, mathematically principled, and experimentally validated method for direct geolocation and cancellation of terrestrial GNSS jammers from single-satellite, multi-antenna LEO platforms under practical SWaP constraints. By compressing and adaptively filtering I/Q data in the time-frequency domain and leveraging position-space cross-correlation approximations, QDG achieves geolocation accuracies within km scales with minimal data volumes. Its flexibility to process-on-orbit or after small downlinks unlocks scalable and persistent global GNSS interference surveillance, laying the groundwork for robust, distributed, and low-latency space-based RFI monitoring networks. Future work is focused on deploying QDG in real-time multi-satellite scenarios and optimizing for more complex interference environments.