EMMA-POS v1

Warning

Pre-implementation. This page describes proposed contracts. Class signatures, parameter types, schema fields, and behavior are subject to change before code lands. Once implementation exists, content here will be regenerated from docstrings or sourced from running tests.

Dataset. EMMA-POS-v1 covers the task, target domain, input, labels, splits, and the download command; this recipe pins only the rfgen configuration. Source. rfgen (primary).

rfgen recipe

  • rfgen commit: <rfgen-commit> (frozen at release; placeholder until pinned).

  • Emitters: the rfgen emitter_zoo group, comms family, placed at known geographic positions across varied angles and ranges. Emitter and receiver placement come from the rfgen scene and placement groups.

  • Receiver array: 8-element URA (uniform rectangular array), half-wavelength spacing (spacing_lambda: 0.5), via the rfgen MultiRXConfig.geometry member URA. Loaded into ArraySpec at load time.

  • Channel environment(s): E_URBAN and E_RURAL, pinned through the rfgen channel group.

  • Labeler + label extraction: the rfgen label group records emitter position; LabelExtractor maps emitters[].position_2d / emitters[].position_3d to the regression target, building on the AoA (angle-of-arrival) metadata from EMMA-LOC v0.1.

  • SNR (signal-to-noise ratio) regime: drawn from the rfgen channel.snr_db_range.

  • Sample counts: <to be pinned at v1>.

  • Splits + seeds: train / dev / holdout seed ranges pinned at release; the holdout is a frozen secret seed, never published. OOD axis: LEAVE_ONE_ENVIRONMENT_OUT (secondary: DEVICE).

Real-capture counterpart

EMMA-REAL-OOD-v0.1 (fidelity reference, carried forward).

Notes

  • Positioning inherits the AoA phase-coherence dependency: a 2D / 3D fix reads the same inter-antenna phase, so the rfgen phase-fidelity verification is load-bearing here too.

  • The layered DEVICE axis probes whether the fix generalizes across receiver hardware, not just across channel environments.

References

  • 3GPP TR 38.843, “Study on AI/ML for NR.” Names positioning as an AI/ML use case without dataset or baseline. (verify revision-year)

  • Schmidt, “Multiple emitter location and signal parameter estimation,” IEEE Trans. Acoustics, Speech, Signal Processing 1986, DOI:10.1109/TASSP.1986.1164830. The MUSIC algorithm; the angle foundation positioning extends. (verify)

See Also