IROS Submission

Final-gate audit for an IROS conference submission on PaperPlaza (ras.papercept.net), the shared RAS system that also runs ICRA, CASE, and RA-L. Everything here is anchored to the verified IROS 2026 cycle; the numbers move each edition because a new IEEE/RSJ committee resets them, so open the current year-site's call-for-papers page before trusting any value below.

Hard limits (2026-cycle values — reverify each year)

The tighter video envelope is an IROS signature: where ICRA allows 180 seconds and 20 MB, IROS 2026 allows only 60 seconds and 10 MB, so a video cut for ICRA must be re-edited, not just re-uploaded.

The references-count-inside trap

IROS counts references (and any appendix) within the page budget. Authors arriving from ML venues that advertise "N pages excluding references" routinely overrun. There is no separate supplementary PDF either — so an argument that does not fit the body does not get an appendix to escape into. Budget the bibliography as content from the first outline, not as free space.

The 2026 anonymity posture

IROS carried a long single-blind tradition, but the 2026 cycle presents as double-anonymous. Concretely:

• Strip the author block, affiliations, and identifying acknowledgements/grant numbers from the PDF; keep the final author list in PaperPlaza metadata (treat it as binding, not a placeholder).
• Rewrite platform tells: "our previously built XYZ rover [9]" becomes "the XYZ rover [9]" cited in the third person.
• De-identify figures and the video: institute banners, lab door signs, logo'd safety vests, and building-specific scenery are all leaks.
• Anonymize links: a code URL carrying a lab organization name is an identity leak; use an anonymized mirror or omit it at review time.
• Because the policy departs from IROS habit, confirm it still holds for your target year before you de-anonymize or re-anonymize anything.

PaperPlaza mechanics worth knowing early

• Every co-author needs a PaperPlaza PIN tied to a profile; gathering PINs from a large author list on deadline night is a known time sink — do it the week the site opens.
• The portal runs its own PDF compliance test (embedded fonts, page size, page count); a failing PDF can be re-uploaded until the deadline, so test days early, not hours early.
• Keywords come from a fixed RAS taxonomy and steer which Associate Editor and reviewer pool see the paper; pick the two or three closest, not the broadest.
• The confirmation email is the only receipt — archive it.

Dual-submission and the RA-L boundary

The same manuscript may not be under review at IROS and elsewhere at once. Note the current-cycle shape:

• IROS 2026 has no combined RA-L+IROS submission option; do not try to submit "to both."
• The legitimate journal-to-conference route is a published RA-L (or T-RO, T-ASE, RAM) paper transferred for presentation during an eligibility window — a different mechanism from submitting.
• A preprint on arXiv is generally allowed; confirm the exact wording in the current CFP.

Pre-upload sequence (final 48 hours)

1. Rebuild the PDF from a clean clone; confirm the IEEE class file is untouched.
2. Page audit: content + references + appendix all fit the current budget; note any page charge.
3. Anonymity sweep: pdfinfo metadata, acknowledgements, self-citations, figure backgrounds,
   embedded links, and every video frame and caption.
4. Play the 60 s video end to end on a machine without lab codecs; verify duration, size, format.
5. Verify PaperPlaza metadata: authors + PINs, keywords, abstract synced to the PDF.
6. Upload paper and video; pass the compliance check; archive the confirmation email.

Output format

[IROS upload verdict] go / fix-first / no-go
[Page audit] content+refs <n>/<limit>, page charge? <y/n>
[Anonymity audit] pass / leaks: <list>
[Video audit] <s>s / <MB>MB / format / deadline ok?
[Metadata audit] PINs collected? keywords chosen? abstract synced?
[Dual-submission] clear / conflict: <venue>
[Deadline math] <local time> vs <CFP zone> = <hours remaining>