Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 9 additions & 0 deletions papers/anom_combined/.gitignore
Original file line number Diff line number Diff line change
@@ -0,0 +1,9 @@
# LaTeX build outputs; the scripts regenerate tables/ and figures/, which are tracked.
main.pdf
*.bcf
*.blg
*.run.xml
*.fls
*.out
*.synctex.gz
__pycache__/
919 changes: 919 additions & 0 deletions papers/anom_combined/IONconf-v2.cls

Large diffs are not rendered by default.

90 changes: 90 additions & 0 deletions papers/anom_combined/NOTES.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,90 @@
# Open items before submission

Status (2026-10-02): first complete draft. 19 pages in the NAVIGATION template, including
references (the limit is 20). 9 figures (limit 15), 9 tables, abstract 267 words.

## Needs an author decision

1. **Correction framing (Section 2).** The paper withdraws the ITM 2026 and Pacific PNT 2026
results explicitly. Agree the wording with Dr. Dames, and decide whether to tell ION about
the conference papers separately (erratum or retraction notice). Under ION's policy a journal
version of a conference paper is normal; a journal version that *reverses* it should say so
in the cover letter as well.
2. **Title.** The working title is descriptive. Alternatives:
"Is a Smartphone Enough? ..." or "Sensor Quality, Not the Estimator, Decides ...".
3. **Abstract length.** 267 words. The NAVIGATION submission page does not give a limit; check
it in ScholarOne and trim if needed.
4. **Experimental-design diagram.** Planned as a TikZ figure and left out to stay under 20
pages. Add it if a reviewer asks, or if the page count drops.
5. **Front matter.** Affiliation, correspondence email (`jbrodovsky@temple.edu` was copied from
the data paper), acknowledgments, and the conflict-of-interest statement. NAVIGATION also
needs an author photo and, on acceptance, a 2–3 minute video abstract.

## Provenance gaps

6. **Smartphone-arm runs.** `data/output_real2` (identical to `data/output_real`, Sep 26
18:31–20:27) has no recorded recipe. `conf/real/rbpf_*.toml` no longer parses with the
current code. Before submission, regenerate the smartphone arm from a committed config set
(e.g. `conf/real/` updated to the Canciani RBPF keys, with paths pointing at
`data/input_real` and `data/output_real`) and the commit that produced `data/output`. Then
rerun `scripts/*.py`. The data-availability section promises that every number can be
regenerated, and right now this one arm cannot be.
7. **Commit of record.** The `data/output` runs (2026-09-26 21:26–23:01) read `conf/` as
committed at `2a40f1e` (09:43 that day). That is where the RBPF recipes became 2,000
particles for full GNSS and 1,000 for degraded and aided; at `827db06` all five said 2,000.
But `target/release/strapdown-sim` in the main checkout is dated 2026-09-25 21:18, before
both commits, so the executing revision is unverified. The paper now says so. Rerunning from
a tagged commit and archiving the binary's `--version` would close this.
8. **Conference-era defects.** Every defect in Section 2 was confirmed by reading the source at
`4a4e3de` (2025-11-14, "Write UKF Anomaly paper") and `4ea73af` (2026-03-11, "anomaly rbpf
paper"). This assumes those are the revisions the conference runs used. The missing
heading update (aided vs. unaided streams) applies only to `4ea73af`: at `4a4e3de` neither
stream had one.

## Numbers that differ from the dissertation (the paper's own are recomputed from the CSVs)

- **UKF α = 0.1**, not 1e-3. `DEFAULT_UKF_ALPHA` was raised in `b70cd55` (2026-09-18), before
the 2026-09-26 runs, and no config overrides it. Dissertation Ch. 3–4 should be corrected.
- **RBPF particles = 1,000** in every config, not 2,000 with full GNSS.
- **Map gradients.** Along-track median 1.44 mGal/km and 6.37 nT/km, from the gradient of
the bilinear cell containing each point (`build_figures.bilinear_gradient`, after PR review),
against 1.3 and 6.9 in the dissertation. Single-update scales: 91 / 1,183 / 38 / 0.6 km;
indicative noise targets 0.6 mGal / 2.8 nT. The paper now presents these as a directional
heuristic, not a localization bound.
- **Drive 2025-06-14.** The current input has 197 fixes, a median reported accuracy of 12.2 m,
and 9 fixes at ≥ 999 m; the dissertation's "587 rows, 16.2 m" is from the 1 Hz pipeline.
- **Gravity-norm vs. map slope.** −0.02 against −0.012 (correlation −0.055 matches).
- **Heading dependence of the magnetic intensity.** 52% of variance from a cos/sin fit to GNSS
course over ground, against 57% in the dissertation from a heading-circle fit. The
dissertation's 56 µT body-fixed field and 6,257 nT 60-s residual were **not** reproduced,
so the paper does not quote them. The 67–325 µT OS hard-iron offset and the
raw-accelerometer result (50–90× the map signal) are cited to the dissertation, not
recomputed.
- **The two unaided sets** differ (EKF mean 611.4 vs. 601.7 m) because rescaling the
magnetometer vector changes the heading input in the last bit. That changes one EKF drive by
261 m and 22 of 25 RBPF drives by >1% (median 20%). The paper reports this. Each arm is
scored against its own unaided runs.
- **Paired statistics.** The aided RMSE comes from `analyze geoperformance`
(`*_detailed_results.csv`). It differs from `analyze performance` by <0.1 m on some drives
(e.g. 759.95 vs. 759.90 m), presumably time alignment. The baseline table uses
`performance`, the ratios use `geoperformance`, and each is internally consistent.
- **Known data-integrity issues, handled.** `data/output/rbpf/{grav,both}/2023-08-04_21-47-58.csv`
are stale leftovers whose baseline failed. The matched-set rule (both runs must exist) drops
them.

## Citations

- `strapdown-rs`: update to the JOSS DOI once joss-reviews#11377 is accepted.
- `mems-nav-dataset`: cited as the Zenodo record 10.5281/zenodo.17582434. That record holds 17
drives at ~1 Hz, while this paper uses 27 segments at 10 Hz. Publish the new release and cite
its DOI.
- `brodovsky2026dissertation`: confirm the title, year and degree status. Check NAVIGATION's
policy on prior appearance in a dissertation (usually not prior publication; say so in the
cover letter).
- `groves_ch5` / `groves_ch14` render as Groves (2013a/b). They are the same book, so cite it
once with chapter pinpoints if the editor prefers.

## Venue facts (ion.org/navi/submit-navi.cfm, read 2026-10-02)

Regular paper ≤ 20 pages, ≤ 15 figures, APA 7 with DOIs, ScholarOne submission. Open access
APC: $1,000 (member) / $1,500 (non-member) for ≤ 20 pages.
48 changes: 48 additions & 0 deletions papers/anom_combined/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,48 @@
# Anomaly aiding: EKF, UKF and RBPF (ION *NAVIGATION* manuscript)

Journal manuscript that combines and **corrects** the two ION conference papers
(ITM 2026, UKF; Pacific PNT 2026, RBPF) and adds the EKF. Template: ION's NAVIGATION
LaTeX template (`IONconf-v2.cls`, from `https://www.ion.org/navi/upload/LaTexTemplate.zip`),
APA 7 via biblatex/biber.

## Layout

| Path | What |
|---|---|
| `main.tex` | Preamble, front matter, section includes |
| `sections/*.tex` | One file per section |
| `tables/*.tex` | **Generated** tables and `numbers.tex` / `phone_numbers.tex` macros quoted in the text |
| `figures/*.pdf` | **Generated** figures (`ion_logo.pdf` comes from the template) |
| `scripts/stats.py` | Loads run outputs; paired ratio, bootstrap CI, Wilcoxon |
| `scripts/build_tables.py` | Writes `tables/` |
| `scripts/build_figures.py` | Writes `figures/` |
| `scripts/phone_mechanism.py` | Writes `tables/phone_numbers.tex` (why the phone channels are uninformative) |
| `scripts/rbpf_diagnostics.py` | Writes `tables/rbpf_numbers.tex` (RBPF crossing time, reported σ vs. error) |
| `build_figures.py` also writes | `tables/bound_numbers.tex` (map gradients, single-update scales, noise targets) |
| `NOTES.md` | Open items for the authors |

## Rebuild

The run outputs are not in git. The scripts read them from `STRAPDOWN_DATA`, which defaults to
`data/` at the root of the checkout that holds the scripts (set it when the data live
elsewhere, e.g. when working in a git worktree):

- `data/output` — dedicated-sensor (synthetic ADXL355/RM3100) arm, plus the unaided runs
- `data/output_real2` — smartphone arm
- `data/input`, `data/input_real` — the preprocessed 10 Hz inputs of each arm

```bash
export STRAPDOWN_DATA=/home/james/Code/strapdown-rs/data
uv run --with scipy python papers/anom_combined/scripts/build_tables.py
uv run --with scipy python papers/anom_combined/scripts/phone_mechanism.py
uv run --with scipy python papers/anom_combined/scripts/rbpf_diagnostics.py
uv run --with scipy python papers/anom_combined/scripts/build_figures.py
cd papers/anom_combined && latexmk -pdf main.tex
```

`build_tables.py` is deterministic (seeded bootstrap): rerunning it on unchanged data gives
byte-identical `tables/`.

The dedicated-sensor and unaided runs read the `conf/*.toml` recipes as committed at `2a40f1e`
(2,000 RBPF particles with full GNSS, 1,000 otherwise); the binary's revision was not recorded. The provenance of the smartphone-arm runs in
`data/output_real2` (which configs, which commit) is an open item: see `NOTES.md`.
Binary file added papers/anom_combined/figures/fig_baselines.pdf
Binary file not shown.
Binary file added papers/anom_combined/figures/fig_bound.pdf
Binary file not shown.
Binary file added papers/anom_combined/figures/fig_example.pdf
Binary file not shown.
Binary file added papers/anom_combined/figures/fig_maps.pdf
Binary file not shown.
Binary file not shown.
Binary file added papers/anom_combined/figures/fig_ratios.pdf
Binary file not shown.
Binary file added papers/anom_combined/figures/fig_rbpf.pdf
Binary file not shown.
Binary file added papers/anom_combined/figures/fig_residuals.pdf
Binary file not shown.
Binary file added papers/anom_combined/figures/ion_logo.pdf
Binary file not shown.
68 changes: 68 additions & 0 deletions papers/anom_combined/main.tex
Original file line number Diff line number Diff line change
@@ -0,0 +1,68 @@
\documentclass[letterpaper,times]{IONconf-v2}

% IONconf-v2 (ION's NAVIGATION template, LaTexTemplate.zip) already loads amsmath, graphicx,
% booktabs, tabularray, tikz, xcolor and biblatex-apa with natbib=true.
\usepackage{bm}
\setcounter{MaxMatrixCols}{20}
\usepackage{siunitx}
\usepackage{algorithm}
\usepackage{algpseudocode}
\usepackage{multirow}
\usetikzlibrary{arrows.meta,positioning,fit,calc}
\sisetup{group-separator={,},group-minimum-digits=4}

\addbibresource{references.bib}

\usepackage[hidelinks]{hyperref}
\usepackage[capitalize,noabbrev]{cleveref}

\graphicspath{{figures/}}

% Numbers quoted in running text, generated by scripts/build_tables.py.
\input{tables/numbers.tex}
\input{tables/bound_numbers.tex}
\input{tables/rbpf_numbers.tex}

\articletype{Original paper}%
\journalname{NAVIGATION}

\title{Gravity and Magnetic Anomaly Aiding of a Smartphone-Grade Inertial Navigation System under Degraded GNSS: A Paired Evaluation of Kalman and Particle Filters}

\author[1]{James Brodovsky*}
\author[1]{Philip Dames}

\authormark{BRODOVSKY \textsc{and} DAMES}

\address[1]{Department of Mechanical Engineering, Temple University, Philadelphia, Pennsylvania, USA}

\corres{*James Brodovsky, Temple University, Philadelphia, PA, USA.\\ \email{jbrodovsky@temple.edu}}

\begin{document}

\abstract[Abstract]{\input{sections/00-abstract}}

\keywords{alternative navigation, geophysical navigation, magnetic anomaly navigation, gravity anomaly navigation, MEMS inertial navigation, particle filter, unscented Kalman filter, GNSS degradation}

\maketitle

\input{sections/01-introduction}
\input{sections/02-corrections}
\input{sections/03-navigation}
\input{sections/04-geophysical}
\input{sections/05-experiment}
\input{sections/06-results}
\input{sections/07-discussion}
\input{sections/08-conclusion}

\section*{Data and Software Availability}
\input{sections/09-availability}

\section*{Acknowledgments}
The authors thank the volunteers who recorded drives for the MEMS-Nav dataset.

\section*{Conflict of Interest}
The authors declare no potential conflicts of interest.

\printbibliography[title=References]

\end{document}
Loading
Loading