# Citation Audit — JPL_AUTONOMY_EDL_03

Read-only audit of the reference list in `dissertation.md` (section `## References`, refs [1]–[128]).
The dissertation is **not** modified by this audit. This file is a repair worklist for Nate.

Audit date: 2026-06-19. Resolver of record: doi.org global handle system (responseCode 1 = live).
Crossref REST API (`api.crossref.org/works/<doi>`) was the first-pass probe; where Crossref
returned 404 the entry was re-checked against the authoritative doi.org handle API and a live
landing-page GET before any "dead" verdict, per the guardrail (a non-Crossref registrant or a
publisher block is *unverifiable*, never *dead*).

## Summary

| Metric | Count |
|---|---|
| Total references | 128 |
| References carrying a DOI | 125 (all unique) |
| References carrying a URL only | 3 (refs [8], [85] Princeton University Press; [109] NTRS) |
| **Resolved (live record confirmed)** | **128** |
| Unverifiable (timeout / 403 / network) | 0 |
| **Flagged dead (confirmed 404/410 handle, or near-miss)** | **0** |

### Counts by classification

| Classification | Count |
|---|---|
| resolves | 128 |
| mistyped_recoverable | 0 |
| misattributed | 0 |
| fabricated_neighbor | 0 |
| not_found | 0 |
| unverifiable | 0 |

**No defects found.** Every DOI resolves at the registry of record, every URL is reachable,
and the NTRS citation resolves via its API endpoint. ARGOS near-miss resolution was therefore
not triggered for any entry (it runs only on confirmed-dead handles). The dissertation's own
back-matter claim in Appendix C — "125 unique digital object identifiers and 128 resolvable
links … each clickable doi.org link … points to a live record" — is corroborated by independent
resolution.

## Resolution method notes (why a naive Crossref-only sweep mis-reports four entries)

A Crossref-only probe flags four entries as 404. None is a real defect. They are recorded here
so a future re-audit does not re-raise them:

| Ref | Cited DOI | Crossref API | doi.org handle | Landing page | Why Crossref 404s |
|---|---|---|---|---|---|
| [11] | `10.1016/0022-4898(72)90071-7` | 200* | LIVE (rc 1) | Elsevier / *J. Terramechanics* | *Tooling artifact only.* The DOI contains literal parentheses; a naive regex that stops at the first `)` truncates the DOI to `10.1016/0022-4898(72` and 404s. The full, correctly URL-encoded DOI returns Crossref 200. No fix needed. |
| [12] | `10.1016/0022-4898(73)90118-3` | 200* | LIVE (rc 1) | Elsevier / *J. Terramechanics* | Same parenthesis-truncation artifact as [11]. Full DOI returns Crossref 200. No fix needed. |
| [16] | `10.1184/r1/6560684` | 404 | LIVE (rc 1) | `kilthub.cmu.edu/.../Results_of_the_inflatable_robotic_rover_testbed/6560684` (HTTP 200) | DataCite (CMU KiltHub/Figshare) DOI, not a Crossref registrant. Crossref's API legitimately does not index it; the handle resolves and the landing page matches the cited title (Apostolopoulos, *Results of the Inflatable Robotic Rover Testbed*). Unverifiable-by-Crossref, **confirmed live**. No fix needed. |
| [78] | `10.82308/1711` | 404 | LIVE (rc 1) | `mcgill.scholaris.ca/items/07df2328-...` (HTTP 200) | McGill institutional-repository DOI (registrant prefix 10.82308), not Crossref-registered. Handle resolves; landing page matches the cited title (Mac Mahon, *Modelling and contact analysis of planetary exploration rovers*). Unverifiable-by-Crossref, **confirmed live**. No fix needed. |

\* Crossref returns 200 only when the DOI is queried in full; the asterisk marks the first-pass
truncation artifact described above, corrected on recheck.

## Flagged references table

None. There are no dead, mistyped, misattributed, or fabricated references to repair.

The 125 DOIs span Crossref (the large majority), DataCite/institutional ([16], [78]), and a
Copernicus/EGU/EPSC abstract series ([6] `egusphere-egu22-7695`, [98] `epsc2024-377`) — all of
which resolve. The three URL-only entries resolve: Princeton University Press book pages ([8],
[85], HTTP 200) and the NTRS citation ([109]) via `ntrs.nasa.gov/api/citations/17627231162643`
(HTTP 200; the NTRS SPA 200s every route, so the API endpoint was used per the hardened rule).

## Provenance caveat carried from the dissertation (not a citation defect)

Appendix C records that two vault literature services were down during corpus assembly
(Semantic Scholar expired host certificate; IEEE Xplore inactive developer key), while OpenAlex,
Crossref, NTRS, and Scopus responded. That is a coverage caveat on the *literature sweep*, not a
defect in any cited entry. Every entry that made it into the list resolves.