docs: sync RTK framing to post-PR #19-#23 demo5 dominance#28
Merged
rsasaki0109 merged 1 commit intodevelopfrom Apr 20, 2026
Merged
docs: sync RTK framing to post-PR #19-#23 demo5 dominance#28rsasaki0109 merged 1 commit intodevelopfrom
rsasaki0109 merged 1 commit intodevelopfrom
Conversation
The README + benchmarks page were still written around the pre-Phase-2 "precision-first equivalent-or-above" framing and an Odaiba-only table. With PR #19–#23 landed, the real picture on production datasets is much stronger — and the README is the first thing people read. - Replace the RTK-vs-demo5 section with three tables: - PPC Tokyo run1/2/3: dominant in Fix count, Fix rate, Hmed, Vp95 at the `low-cost` preset with no Phase 2 flags. - PPC Nagoya run1/2/3: same preset, same picture. - UrbanNav Odaiba: dominant on Fix count, Hp95, Vp95 by default; the one demo5 edge (Hmed) closes to 9 cm with `--enable-wide-lane-ar --wide-lane-threshold 0.10`. - Add a new "Phase 2 opt-in tuning gates" section listing the five default-off flags introduced in PR #19–#23, with a note that they are byte-identical to the default unless explicitly enabled. - Update the `Benchmark Snapshot` Odaiba detail block to the current numbers (the old `11637 / 8241 fixed`, `Hmed 0.733 m / 0.704 m` snapshot is from an earlier develop). - Mirror the multi-dataset table into `docs/benchmarks.md`. - In `docs/references/demo5-gap.md`, note that the `arfilter` row now also has `--ar-policy demo5-continuous` (PR #19), and that the hold row is covered by `--max-hold-div` (PR #20) and `--max-postfix-rms` (PR #22) except for the `varholdamb`/`gainholdamb` semantic slice. All PPC/Nagoya/Odaiba numbers were reconciled against a fresh `apps/gnss.py ppc-demo` run on this branch's build (post-PR #27 develop) using `--preset low-cost --match-tolerance-s 0.25`; all six runs match the numbers written here. No code changes; `docs/decisions.md`, `docs/architecture.md`, `docs/quickstart.md`, and the other `docs/references/*-gap.md` files are intentionally untouched.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
README + benchmarks page were still on the pre-Phase-2 "precision-first equivalent-or-above" framing and only showed an UrbanNav-Odaiba table. With PR #19–#23 landed the real picture on production datasets is stronger — and the README is the first thing people read.
Reconciliation
All six PPC-Dataset numbers were regenerated on this branch's build (post-PR #27 `develop`) with `apps/gnss.py ppc-demo --preset low-cost --match-tolerance-s 0.25 --max-epochs -1`. Every row in the README / benchmarks.md tables matches the regenerated `summary.json` output:
UrbanNav-Odaiba numbers match `output/benchmark/odaiba/default_vs_rtklib.json` and `wl010_vs_rtklib.json` (both 2026-04-20).
Touched
Not touched: `docs/decisions.md`, `docs/architecture.md`, `docs/quickstart.md`, `docs/references/*-gap.md` (other gap docs are per-reference analysis, not benchmark framing).
Test plan