Satellite License Plate: passive and compact optical spectrally-based identification method for satellites

The study addresses the growing need for reliable satellite identification amid rapidly increasing launches and congested orbits, where potential collisions and abundant unidentified debris complicate space traffic management. Current identification relies on radar and passive optical observations, which are insufficient for uniquely identifying many LEO objects—particularly individual spacecraft within swarms before radio contact is established. Cooperative approaches that place hardware on satellites can be active (e.g., ELROI using powered laser diodes) or passive. Active methods require continuous power, including backup for non-operational spacecraft, which is burdensome for small satellites. Passive methods leverage retroreflectors and unique tag properties (e.g., number of retroreflectors, polarization) but can demand significant surface area or complex ground equipment. This work proposes SLP: a passive, spectrally encoded tag read by multi-wavelength lasers and spectrally resolved receivers, enabling parallel interrogation of channels during the full visibility window. The paper details the SLP concept, decoding approach, end-to-end LEO feasibility, and ground-to-ground validation.

Existing identification methods include radar and passive optical techniques that infer identity via trajectory and light-curve signatures, enabling limited cooperative and non-cooperative identification. Cooperative approaches add satellite hardware for direct identification. Active systems like ELROI (optical license plates using laser diodes) can support many IDs but require power and backup supplies, limiting utility if a satellite fails. Passive cooperative methods use retroreflective tags, with IDs encoded via number of retroreflectors or polarization states; the former scales poorly with surface area, and the latter requires more complex ground polarization handling. Spectral encoding with corner cube retroreflectors (CCRs) and band-pass filters (BPFs) offers orthogonal channels with simpler parallel spectral multiplexing/demultiplexing at the ground station, reducing system complexity and potentially increasing data by using the entire pass duration without time-division interrogation.

Concept and tag: SLP tags are arrays of CCRs paired with BPFs at distinct wavelengths. A reference CCR without a filter provides a wavelength-independent radiometric reference. Illumination with a multiplexed, multi-wavelength laser beam yields retroreflections whose intensities per channel scale with the number of CCRs behind BPFs at those wavelengths. The spectral signature is formed from ratios of received channel intensities, enhancing robustness to overall power fluctuations. A dictionary of valid tag words is defined by the total number of CCRs N (including one reference) and the number of spectral channels L. The number of unique identifiers grows combinatorially with N and L.

Optical Ground Station (OGS): A narrow-FOV telescope tracks the satellite using ephemerides refined by passive optical tracking. The transmitter combines multiple spectrally spaced lasers (tens of nm apart) through collimation, multiplexing, and a beam launcher (afocal telescope). The receiver collects the return, demultiplexes spectrally, and detects with per-channel photodetectors (e.g., APDs), then digitizes and decodes via maximum-likelihood (ML) matching of the measured signature to the dictionary signatures. A bi-static (or mono-static in the field test) assembly ensures separation of Tx/Rx paths while maintaining alignment.

Channel design and BPF angular response: BPFs shift with angle of incidence (AOI), producing a blue-shift and possible inter-channel cross-coupling off-axis. Channel spacing must be sufficient to keep out-of-band transmission negligible at expected AOIs. Example: for BPFs with effective refractive index ~2.5 and 1/e half-bandwidth ~5 nm, a 15 nm spacing at 30° AOI limits cross-coupling below −72 dB. Excessive total bandwidth increases transmitter complexity; practical systems keep the span within tens to hundreds of nanometres.

End-to-end LEO simulation: Two-stage model. (1) Orbit and visibility: Propagate a circular LEO at ~500 km altitude, inclination ~82.9°, OGS at 52.1098 N, 4.3275 E. Determine visibility windows considering geometry, tag angular acceptance, and OGS elevation limits. Assume a nadir-pointing satellite with 4 lateral faces; define visibility when at least one face AOI allows acceptable return (e.g., <~20° AOI yields ~6 dB CCR cross-section loss). Example visibility window is ~45 s. (2) Link and detection: Simulate time-event propagation of pulsed lasers (per channel: ~100 μJ pulse energy, ~10 ns duration, 10 Hz repetition; channels across ~1535–1580 nm). Compute return using CCR array far-field diffraction, summing single-CCR patterns incoherently. Model per-channel detection chains: APD, CTIA amplifier, ADC, including electronic noise and night-sky background spectral radiance. Form measured spectral signatures as ratios across channels and apply ML decoding by minimizing distance to nominal dictionary signatures (computed assuming high in-band BPF transmission and negligible cross-coupling). For a test case with N=5 CCRs (1 reference) and L=4 channels, the signature lies in a 6D ratio space and the dictionary has 35 tag words.

Ground-to-ground free-space experiment: Purpose: feasibility demonstration and characterization of identification and angular response using only off-the-shelf components. Geometry: 2.43 km one-way free-space path in The Hague (NL). OGS transceiver (mono-static) on a tower at ~37.5 m elevation; tag at ~40 m elevation at a separate site. Tag: array of 5 CCRs; two BPF types: 1540 nm and 1560 nm; dictionary defined by L=2, N=5 → 5 unique IDs plus an unfiltered reference configuration for calibration. Transmitter: DWDM source with two DFB lasers (1540 and 1560 nm) combined with OADMs, intensity-modulated by an EOM driven by an AWG (with RF amplification), amplified by EDFA, polarization conditioning, and launched through a beam expander (10×) with slight divergence (half-angle ~70 μrad, beam radius ~45 mm at aperture). Receiver: spectral separation via PBS and dichroic/filter chain to two APD detectors (per channel), digitized by oscilloscope; FSM available for fine pointing (used open-loop for alignment). Procedures: coarse visual alignment followed by fine alignment using scanning with a large retroreflector; then switched to the CCR tag. Modulation: 5 MHz square-wave bursts, 50% duty cycle; Round A: 16 μs on, Round B: 8 μs on, followed by 32 μs off; oscilloscope averages 2048 traces per recorded waveform. Dictionary runs interrogated sequences: {00} (no BPFs, calibration), then {40}, {31}, {22}, {13}, {04}; Round A collected 20 averaged waveforms per configuration; Round B collected 100 traces per configuration.

Data processing and decoding: For each waveform and channel, extract time windows of near-field back-reflection and delayed retroreflection. Fit and subtract a 3rd-order polynomial to mitigate slow drifts. Compute RMS of the detrended signals. Form normalized signatures by dividing retroreflected by back-reflected RMS per channel to compensate temporal channel fluctuations. Correct residual inter-channel imbalance (e.g., EDFA gain curve) using calibration factor derived from the {00} measurements. Compare resulting signatures to nominal dictionary via ML decision; summarize performance with confusion matrices and Matthews Correlation Coefficient (MCC). Angular response test: with tag {22}, rotate azimuthally in 2° steps, acquire normalized return vs AOI; compare to models with CCR geometric cross-section loss alone and with additional BPF AOI-dependent transmission loss (fitted n_eff≈2.5, BW≈10.7 nm).

• End-to-end LEO simulation (N=5, L=4; 35 tags): ML decoding achieved a 97% correct identification rate (1 mis-decoded tag out of 35) under modeled night-time conditions, with a representative visibility window of ~45 s. Nominal channels designed with ≥15 nm spacing at ~30° AOI kept cross-coupling below ~−72 dB for BPF BW≈5 nm and n_eff≈2.5.
• Ground-to-ground dictionary tests (2.43 km, two wavelengths at 1540 and 1560 nm, N=5): Distinct, monotonic signature trends matched the theoretical linear behavior as the number of BPFs per channel varied, demonstrating feasibility of spectral-ID readout. However, measurement spread exceeded the ideal Voronoi regions for unambiguous classification, reflecting channel fluctuations and system drifts. Confusion matrices yielded MCC≈0.55 (Round A) and ≈0.28 (Round B), indicating strong and weak positive predictive correlations, respectively. Removing the most confusable tag configurations ({31} and {13}) improved robustness (per supplementary analysis).
• Angular response: Measured normalized return versus AOI aligned well with a model incorporating both CCR geometric loss and BPF angular transmission dependence (fit parameters n_eff≈2.5, BW≈10.7 nm). A geometric-only model underestimated the loss, confirming the necessity of including BPF AOI effects in design and performance prediction.
• System observations: Experimental necessity for inter-channel normalization and calibration arose from EDFA gain non-uniformity and free-space/polarization drifts; the adopted normalization and calibration using {00} data successfully mitigated these effects for analysis.

The findings support the central hypothesis that a passive, spectrally encoded retroreflective tag can enable unique satellite identification from ground-based optical stations. Simulations indicate high decoding accuracy with realistic laser powers, detectors, and spectral channel design. Field tests over multi-kilometer paths verified that spectral signatures can be measured and mapped to IDs, although real-world variations (atmospheric turbulence, pointing, polarization and amplifier drifts) broaden signature distributions and can reduce single-pass decoding reliability. The angular response study confirms that BPF AOI behavior significantly affects link budget and visibility windows; accurate modeling is necessary for scheduling interrogations, especially for tumbling or non-operational objects where time-varying AOI modulates returns. Practical improvements—such as an in-line calibration branch to track transmitter power and spectral balance in real time, longer averaging over multiple passes, and careful channel spacing—can substantially enhance robustness. Trade-offs between dictionary granularity (ID density) and noise tolerance are evident; coarser coding (larger steps in BPF counts per channel) enlarges decision regions at the cost of fewer unique IDs. The approach is compatible with existing OGS services (SLR, optical comms/observations) and can be combined with orbital information to expand effective ID space (reusing the same tag code on different orbits).

This work introduces and validates the Satellite License Plate (SLP), a passive, compact, spectrally encoded retroreflective tagging method for cooperative satellite identification. The concept leverages multi-wavelength laser interrogation and spectral demultiplexing to read unique tag signatures in parallel over the full pass duration. End-to-end modeling for LEO predicts high identification accuracy with realistic components, and free-space experiments over 2.43 km demonstrate feasibility and inform practical design considerations (e.g., calibration, channel spacing, and AOI effects). SLP integrates naturally with existing OGS infrastructure, enabling potential multi-service operation (e.g., SLR and attitude characterization). Future work should implement in-line calibration paths, explore increased CCR aperture or reduced dictionary granularity to enlarge decision margins, optimize spectral channel design against AOI-induced BPF shifts, and validate performance in true ground-to-space trials. Combining spectral IDs with orbit discrimination can scale identification across large satellite populations, including non-operational objects.

• Experimental demonstrations used only two spectral channels and a small dictionary (5 IDs), limiting direct extrapolation to larger codebooks.
• Significant measurement spread due to atmospheric turbulence, pointing stability, polarization changes, and amplifier gain drift reduced single-pass identification reliability; real-time calibration was not implemented on-line.
• Simulations assumed idealized BPF responses with negligible cross-coupling and did not include BPF AOI loss in the link budget stage (though analyzed separately), potentially overestimating performance margins.
• The repetition rate in simulations (10 Hz) was limited for computational reasons; higher rates could affect averaging behavior and noise statistics in practice.
• Ground-to-ground tests may not fully capture space-to-ground dynamics (e.g., satellite motion, spin states, varying AOI and Doppler), though angular response was partially characterized.
• Off-axis interrogation reduces cross-section; angular acceptance and tag placement on spacecraft faces constrain visibility windows and identification opportunities.