Overview
# 📡 ReadyGary — 6G Beam Selection
7GC · STAGING_WORKER
ReadyGary beam selection
Research repository for AI-native 6G beam selection. Results stay inside the repository's own evidence.
Explore evidence
Evidence class: SYNTHETIC_SIM. The selector lists synthetic beam-selection records. Toy gains are not calibrated mmWave measurements.
Overview
Status levels (honest)
28 GHz is **FR2 mmWave** (3GPP TS 38.101-2), **never Sub-6**. Twin/sim ≠ OTA. `<1 ms` inference is a **TARGET**, not a measured fact. Primary **true below-6 GHz** profile is **n77 3.75 GHz FR1** (Quectel RM520N-GL BOM, US C-band / Gary scenario). FR2 28 GHz remains a separate experiment family. Not every FR1 band is below 6 GHz (n96 is documented as excluded). Reproduce: `make reproduce` · `make sub6-reproduce` · `ma
🔬 **What is ReadyGary?**
ReadyGary is a **research project** studying the critical beam selection problem in 5G/6G mmWave networks. By combining physics-based baselines with learned ML models, we **target** fast inference (measured proof pending — see evidence matrix) for real-time beam tracking under mobility and blockage. **🎓 Based on ECE-6023 Final Project** with comprehensive improvements addressing professor feedback on realistic channe
Methods
Methods and the toy scores are copied from results/benchmark_summary.json and the train policies in results/experiments/rq2_beam_selection_fr2_summary.json. evidence_class is SYNTHETIC_SIM. sub_ms_inference_proven is false. The benchmark note says these are not calibrated mmWave measurements.
Native research runtimes stay on a local checkout. This page shows repository records only.
Artifacts
results/benchmark_summary.jsonresults/benchmark_summary.mdresults/benchmark_table.mdresults/campus_radio/gary_radio_profile_report.mdresults/campus_radio/gaza_radio_profile_report.mdresults/campus_radio/geelong_radio_profile_report.mdresults/campus_radio/germany_radio_profile_report.mdresults/campus_radio/ghana_radio_profile_report.mdresults/campus_radio/graham_land_radio_profile_report.mdresults/campus_radio/guyana_radio_profile_report.mdresults/e2e/README.mdresults/e2e/beam_selection_research_card.md
Source
Pinned source e04b38b48cd2ebd87243153ca8fad9e914f16b21
Limitations
- Not field-validated
- Not carrier-accepted
- Not peer-reviewed by this page
About / Build status / Evidence
What is this?
Research repository for AI-native 6G beam selection. Results stay inside the repository's own evidence.
The research code is not executing here. This is a research surface.
Why does it exist?
It gives this repository a public web surface inside the gunnchOS ecosystem without pretending the original runtime runs on Cloudflare Workers.
What can I actually do with it today?
Use the primary action to inspect repository records, then open the source for the local toolchain.
What is not finished?
- Not field-validated
- Not carrier-accepted
- Not peer-reviewed by this page
STAGING_WORKER
Runtime mode on this page: PRECOMPUTED. No remote model call is made.