Agent Skills
› brycewang-stanford/Awesome-Journal-Skills
› icra-topic-selection
icra-topic-selection
GitHub用于指导机器人项目选择投稿 venue。通过本体验证和贡献类型评估判断是否适合 ICRA，并决策直接投 ICRA 还是通过 RA-L 期刊转会议，避免误投非机器人领域或理论性过强的会议。
Trigger Scenarios
决定机器人项目的投稿目标
在 ICRA 与 RA-L/其他会议间做策略选择
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill icra-topic-selection -g -y
SKILL.md
Frontmatter
{
"name": "icra-topic-selection",
"description": "Use when deciding whether a robotics project belongs at ICRA, choosing between a direct ICRA submission and the RA-L journal route with conference presentation, or re-routing to IROS, RSS, CoRL, T-RO, HRI, or CASE based on contribution type, evidence shape, and the embodiment test."
}
ICRA Topic Selection
Route a robotics project before any formatting work begins. ICRA — the IEEE International Conference on Robotics and Automation, the flagship conference of the IEEE Robotics and Automation Society (RAS) — is the broadest and largest robotics deadline of the year, but "broad" does not mean "anything with a robot picture." This skill runs two decisions in order: (1) is the contribution ICRA-shaped at all, and (2) if yes, should it travel as a direct conference paper or as an RA-L journal paper carrying a conference-presentation option.
The embodiment test
Ask what breaks if the physical robot is deleted from the paper.
- If the answer is "nothing — the method is really about images, language, or generic optimization," ICRA reviewers will say so. Route to a vision, ML, or optimization venue and cite robotics as an application.
- If the answer is "the whole contribution — the method exists because of contact, dynamics, sensing noise, actuation limits, real-time constraints, or task resets," the project is in ICRA territory.
- Middle cases (learning method evaluated only in simulation) can survive at ICRA,
but the burden shifts to the experiments plan; see
icra-experimentsbefore committing.
Fit by contribution type
| Contribution type | ICRA fit | Stronger alternative to weigh |
|---|---|---|
| New manipulation / locomotion / navigation algorithm with hardware trials | Core | RSS if the algorithmic idea is the star |
| Integrated robot system solving a hard task end to end | Core | IROS, field-robotics journals |
| Robot-learning method, sim-heavy, some real transfer | Good | CoRL if learning is the entire story |
| Control / estimation theory with robot validation | Good | T-RO or RA-L for longer proofs |
| Dataset or benchmark for robot tasks | Good | RA-L, IJRR for archival datasets |
| Human-robot interaction with a user study | Possible | HRI conference for study-centric work |
| Factory / logistics automation, scheduling | Possible | CASE, T-ASE |
| Perception with no downstream robot task | Weak | CVPR/ICCV, then a robotics follow-up |
| Pure RL / optimization theory, no embodiment | Weak | NeurIPS, ICML, L4DC |
Direct ICRA vs RA-L with conference presentation
The RAS ecosystem offers two distinct roads to an ICRA podium, and choosing between them is a real strategic decision, not a formality:
- Direct ICRA submission. One deadline per year (mid-September for a late-May/June conference), conference peer review, proceedings in IEEE Xplore, 6 pages plus up to 2 pages of references at submission in the 2026 cycle.
- RA-L (IEEE Robotics and Automation Letters). Year-round submission, first decision targeted within about three months, revise-and-resubmit possible, 6 pages including references with up to 2 extra pages at USD 175 each. Accepted RA-L papers can be transferred, within the posted window and within 270 days of acceptance, for presentation at an upcoming RAS conference such as ICRA — a journal publication plus the same stage.
Decide with these levers:
- Clock. Missed September or the result matures in November? RA-L is the only road that still reaches the next ICRA stage.
- Revision tolerance. RA-L allows a revision cycle; direct ICRA is accept-or-reject on one shot. Work with one fixable weakness favors RA-L.
- CV target. Some fields and hiring committees weight the journal line (RA-L) differently from the conference line (ICRA proceedings). Check what the career context rewards; the presentation looks identical at the venue.
- Length. Both cap near 6-8 pages, so length rarely decides; T-RO or IJRR is the route for genuinely long papers.
- Dual-submission trap. The same manuscript must never sit in the ICRA conference pipeline and at RA-L simultaneously; that is a rejection offense in both systems.
Quick routing script
Q1 Embodiment: does deleting the robot delete the contribution?
no -> route to the parent field's venue; stop.
Q2 Maturity: results solid by early September?
no -> RA-L (rolling) or next IROS deadline (typically March).
Q3 One clean contribution statable in 6 pages?
no -> T-RO / IJRR long-form, or split the project.
Q4 Is the star the algorithmic idea (RSS), the learning method (CoRL),
the user study (HRI), or the factory floor (CASE)?
yes -> weigh that sibling before defaulting to ICRA.
Q5 Otherwise -> direct ICRA; record why in one sentence for the intro.
Lighter tracks at the same conference
Not every result deserves — or needs — the full-paper gauntlet. The ICRA weeks also host workshops with their own paper calls, late-breaking/abstract-style contributions, competitions, and demo slots, each with separate winter/spring deadlines and lighter (often non-archival) review. Correct uses:
- A promising-but-thin result: workshop now, full paper next September with the workshop feedback baked in.
- A dataset or tool needing community exposure before the archival paper.
- A negative or replication result that the main track rarely rewards but a topical workshop will discuss seriously.
Check archival status per call: some workshop papers preclude later "new" submission of the same content, most do not — read the specific call's wording.
Anti-patterns that waste a cycle
- Submitting a workshop-depth demo as a full paper: ICRA has late-breaking, poster, and workshop tracks better suited to preliminary results — check the current year's calls.
- Choosing ICRA purely for prestige when the reviewer pool for the subfield (e.g., soft robotics, surgical robotics) is deeper at a specialized venue.
- Planning "we will add real-robot results during rebuttal" — the classic ICRA
pipeline has no author-response phase to add them in (see
icra-review-process), so the evidence must be complete at submission. - Ignoring that ICRA 2026 moved to double-anonymous review: a framing that leans on "our lab's famous platform" no longer helps and may leak identity.
Output format
[Embodiment test] pass / borderline / fail — <one line>
[Route] direct ICRA / RA-L + presentation option / sibling venue <name>
[Deadline reality] <next reachable deadline and date to verify>
[Contribution type] algorithm / system / learning / theory / dataset / HRI / automation
[Single biggest routing risk] <one line>
[Verify before commitment] current CFP at the ICRA year-site + RA-L information-for-authors page
Version History
- 9f86f09 Current 2026-07-19 15:58
