Code-independent navigation using Signals of Opportunity (SoOP) from non-cooperative low-Earth-orbit (LEO) satellites demands real-time knowledge of each transmitter’s state. In Frequency-of-Arrival (FoA) methods, Doppler changes rapidly because LEOs move at ~7–8 km/s; a few hundred milliseconds of latency or timestamp error can mis-map frequency to line-of-sight velocity and bias positioning. Short visibility windows (tens of seconds to a few minutes), frequent handovers across many satellites, and dense urban or aviation scenarios further compress the processing budget. Since non-cooperative LEOs do not broadcast navigation messages, the receiver must autonomously determine who is visible and propagate state as measurements arrive. Consequently, fast visibility extraction and timely ephemeris propagation are not conveniences; they are prerequisites for FoA observability and for scheduling resources (acquisition windows, handovers, antenna/beam steering, and per-satellite processing).
We present a simple, robust infrastructure that couples a GPU-accelerated visibility engine with a request-driven CPU ephemeris service. A vectorized SGP-4 implementation, validated to sub-millimeter agreement with MATLAB’s SGP-4 over representative horizons, runs on a low-cost NVIDIA Jetson Orin Nano. For merged TLE catalogs exceeding 8,000 LEO spacecraft, the GPU path delivers ~30× speedup over CPU baselines and completes global visibility checks in ≈1 min at sustained 100% GPU utilization; to demonstrate headroom, 50,000 TLE entries are processed in ≈3 min using the same workflow. A lightweight two-layer client–server design (stateless query handler over a pinned-memory ephemeris cache) returns per-satellite state vectors in ≈0.01 s worst-case on CPU while occupying <10% CPU. Under current traffic (~600 simultaneously visible LEOs over a 10-min and without elevation masking), per-request latency averages ≈0.003 s, meeting real-time FoA needs with zero operator interaction and scaling comfortably beyond six times today’s LEO counts. The end-to-end system has passed long-duration stability tests in the LASSENA laboratory; visibility lists and propagated ephemerides are numerically consistent with reference SGP-4 outputs. This architecture offers a practical, commodity-hardware path to autonomous, real-time SoOP navigation at constellation scale.

We present a reproducible workflow to synthesize and exploit non-cooperative LEO Signals of Opportunity (SoOP) in Skydel for Doppler-based navigation, and we validate performance against real-world data under matched satellite geometry. Using direct signal-generation method, each LEO spacecraft is represented as a moving emitter whose trajectory is driven by 1 Hz SGP-4 ephemerides; per-satellite custom baseband IQ is injected and up-converted by Skydel to the RF output. The Iridium-NEXT simplex waveform is emulated using whitening, DE-QPSK mapping with a 125-bit BPSK preamble for fast acquisition, dual-stage RRC pulse shaping, and sub-band translation to the 31.5 kHz channels (41.667 kHz spacing), with multi-channel summation and randomized duty-cycle bursts to mirror TDMA behavior. The same constellation set and site are used for simulation and field tests (L-band center ≈ 1.62627 GHz, sample rate 1 MHz); in replay, eight of nine links are reproduced (one below the elevation mask), yielding Doppler tracks within ±15 Hz (σ≈10 Hz) of predicted S-curves, while over-the-air captures exhibit 50–200 Hz variability attributable to oscillator instability, multipath, and propagation dynamics.
A common processing chain - including coarse acquisition, carrier/Doppler tracking with CFO drift compensation, outlier-robust pre-filtering, and an EKF-based Doppler-only geometry solver - operates identically on simulated and live data. Under identical conditions, Skydel-based navigation achieves ~12 m horizontal error, whereas the real-world experiment yields ~33 m. The gap is consistent with live-environment impairments (residual oscillator biases, local multipath, ephemeris/timing mismatches). Overall, Skydel-generated non-cooperative LEO SoOP faithfully reproduces the key observables and error modes required for design-space exploration and provides credible lower-bound accuracy predictions for resilient PNT when GNSS is degraded or denied.

The rapid development of mega-constellations in low Earth orbit (LEO) has opened up new opportunities for alternative and resilient sources of positioning, navigation and timing (PNT). This paper investigates the feasibility of exploiting SpaceX’s Starlink signals as opportunistic navigation signals, building on the pioneering work of Kassas and colleagues. We focus on the use of Starlink downlink synchronization sequences—specifically the Primary Synchronization Signal (PSS) and Secondary Synchronization Signal (SSS)—to enable user terminals to perform reliable signal acquisition, Doppler estimation, and pseudorange extraction without cooperation from the provider. As comparison and approximate values, the tones in the center of the channel can be used. For assessing feasibility, a hardware setup, composed of a rotary antenna, an SDR (software-defined receiver) and a reference clock have been used. The paper describes the test setup. Furthermore, we evaluate multi-satellite geometry and discuss achievable positioning performance when fusing PSS/SSS-derived measurements. Theoretical and practical investigations are presented. Focus lies on the potential of longer tracked recordings using the rotary antenna. Further, the downsides of not capturing the full bandwidth and clock errors and orbital errors are discussed. Experimental results using over-the-air Starlink recordings demonstrate the achievable pseudorange and Doppler accuracy across diverse orbital passes. The findings show that Starlink-based navigation can serve as a complementary PNT (Positioning, Navigation, and Timing) source, offering enhanced availability, redundancy, geometry and resilience in contested or GNSS-degraded environments.