iros-topic-selection
GitHub辅助判断项目是否适合投稿IROS，指导在ICRA、RSS等会议间路由选择。通过识别集成系统贡献与具身约束，优化智能机器人系统框架，提升论文定位准确性。
Trigger Scenarios
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill iros-topic-selection -g -y
SKILL.md
Frontmatter
{
"name": "iros-topic-selection",
"description": "Use when deciding whether a project is a strong IROS fit, routing among ICRA, RSS, CoRL, RA-L, T-RO, HRI, and CASE from IROS's seat on the fall IEEE\/RSJ calendar, identifying the integrated-system contribution and its embodiment constraint, and sharpening the intelligent-robots-and-systems framing before writing begins."
}
IROS Topic Selection
Use this before writing. IROS is strongest for an integrated intelligent system — perception, estimation, planning, learning, or control working together on a robot — validated under a real embodiment constraint. The house question is not "is the idea novel?" but "does the system run, on stated hardware, and is that demonstrably hard?"
Fit test
- Prefer IROS when the contribution is a working system for perception-action integration, autonomy, navigation, manipulation, robot learning deployed on hardware, human-aware robotics, or fielded systems — and the embodiment (sensors, compute, power, latency) shapes the design.
- Route to ICRA when the work is a fine automation/robotics contribution whose natural slot is the spring cycle; IROS and ICRA overlap heavily, so timing and program fit often decide.
- Route to RSS when the contribution is a single, foundational, science-first claim that wants a single-track audience and a rebuttal, rather than deployed-system breadth.
- Route to CoRL when the learning method is the contribution and robots are the testbed, more than the system integration being the point.
- Route to RA-L / T-RO when the work wants journal review and revision, then optionally a conference presentation via the transfer window.
- Route to HRI / RO-MAN for interaction studies with human-subjects methodology, and to CASE for automation-science and factory-scale problems.
Fit signal table
| Signal in the project | IROS reading |
|---|---|
| A system that runs on real hardware under a compute/power/latency budget | Core fit — the house genre |
| Perception + planning + control integrated and evaluated end to end | Core fit |
| A single foundational theorem or model, science-first, best with rebuttal | Better served at RSS |
| A learning algorithm whose novelty carries the paper, robot as testbed | Better served at CoRL |
| A polished demo with no trials, resets, or failure analysis | Not yet IROS-shaped — build evidence |
| Journal-length exposition or need for revision cycles | RA-L or T-RO, then transfer to present |
Vignette: where a legged-locomotion project goes
A project delivers a controller that keeps a quadruped upright on loose gravel, on the robot, with trial counts and a failure taxonomy. IROS reading: strong fit — an integrated system meeting a terrain constraint with embodied evidence. Strip the hardware and keep only a simulation policy-learning result, and it drifts toward CoRL; distill it to one provable stability claim wanting single-track scrutiny, and RSS becomes the better home; grow it into a journal-length treatment with revisions, and RA-L or T-RO fits, after which the transfer window can still put it on an IROS stage.
Sharpening moves before committing
- Name the embodiment constraint the system respects — onboard compute, power budget, sensor noise, contact uncertainty, or real-time latency. If nothing physical constrains the design, the IROS framing is thin.
- Confirm the argument fits the body alone: IROS has no supplementary PDF, so a system that only makes sense with an appendix is mis-sized.
- Verify you can produce embodied evidence — real trials, resets, failures, and a stated sim-to-real gap — not just a curve or a single hero run.
- Robotics program emphasis drifts between cycles and cities; scan the current CFP topic list before final routing.
Quick router:
system on real hardware under a physical constraint? ---> IROS (or ICRA by season)
single foundational science claim, wants rebuttal? ----> RSS
the learning method is the contribution? -------------> CoRL
wants journal review + revisions? --------------------> RA-L / T-RO (then transfer to present)
Output format
[Fit] strong IROS / possible IROS / better elsewhere
[Best venue] IROS / ICRA / RSS / CoRL / RA-L / T-RO / HRI / CASE / other
[System sentence] <one sentence naming the integrated system>
[Embodiment constraint] <compute / power / latency / contact / sensing>
[Top rejection risk] <evidence / integration / novelty / scope>
[Next action] <build evidence, reframe, or switch venue>
Version History
- 9f86f09 Current 2026-07-19 16:05
