“Find your passion” is not a research method. A useful idea comes from connecting a question that matters with evidence you can realistically gather.
Generate ideas through five doors
1. Start from literature
Find a recent review or survey paper in a field you can name. Use it as a map, not an answer key.
- What does the field mostly agree on?
- Which results conflict?
- What limitations recur across papers?
- Which assumptions fail in unusual populations, environments, scales, or edge cases?
- What future work do multiple authors independently point toward?
- Which dataset, method, or comparison has not been tested in a nearby context?
Follow important references backward. Use “cited by” to find what happened afterward. Read several papers before deciding that a gap exists; one abstract cannot establish novelty.
2. Start from your interests
List what you voluntarily read, build, repair, watch, or argue about. Then move below the topic label:
- What mechanism makes it work?
- What part is inefficient, unreliable, or poorly understood?
- What could be measured rather than merely discussed?
“I like robotics” is a direction. “Why does this navigation method fail near moving people?” is the beginning of a project.
3. Start from everyday friction
Notice repeated workarounds. What wastes time, fails silently, costs too much, excludes someone, creates risk, or produces inconsistent results? Turn the frustration into either a scientific question or an engineering requirement.
4. Start from potential impact
Identify who experiences the problem and how often. Ask what changes if the project succeeds, how that change could be measured, and whether your proposed solution is actually preferable to the existing one.
5. Start from accessible resources
A strong project does not require a university laboratory. Search for public datasets, simulation, inexpensive sensors, open-source software or hardware, observational data, computational models, and equipment your school already owns. Let access constrain the question early—not after months of planning.
Robots
Navigation
Dynamic obstacles
Does adding predicted pedestrian motion reduce collision rate versus a reactive baseline in crowded simulations?
Score candidates before falling in love
Give each idea 0, 1, or 2 points for each criterion. A low score is not a failure; it tells you what must change.
| Criterion | Ask |
|---|---|
| Interesting to me | Will I still care after the easy novelty fades? |
| Useful to someone | Can I name who benefits and how? |
| Measurable | What observation could prove me wrong? |
| Novel enough | Is there a new dataset, comparison, population, mechanism, method, replication, or implementation? |
| Feasible | Can I complete a meaningful version with my time, skills, access, and budget? |
| Safe and allowed | Are safety, privacy, ethics, school, and competition rules satisfied before work begins? |
| Interpretable | If the metric changes, will I know why—or only that it changed? |
| Next experiment | Can I name a one-week pilot? |
“Novel” does not have to mean no human has ever considered the subject. A careful replication under new conditions, a meaningful comparison, an open dataset for an overlooked population, or a simpler implementation can be a real contribution when the field values it.
Two examples
Scientific path
- Broad: indoor air quality
- Narrower: classroom carbon dioxide
- Mechanism: occupancy and ventilation
- Question: “How does opening one versus two exterior doors affect the time required for classroom CO₂ to return below a project-selected comparison level, while outdoor conditions are recorded?”
Engineering path
- Broad: accessible lab tools
- Need: visually aligning a low-cost microscope is difficult
- Constraint: printable parts and a phone camera
- Problem: “Design a phone-microscope alignment fixture under $20 and compare setup time and image repeatability with a hand-held baseline.”
Neither is automatically safe, novel, or competition-ready. Each is specific enough to investigate those questions next.
End with a one-week pilot
Write the smallest experiment, dataset check, simulation, or prototype that could reveal a fatal problem. The pilot should test access, measurement quality, runtime, and interpretability—not prove the final claim.
Afterward, decide: continue, narrow, change method, or stop. Ending a weak idea early is good research judgment.
Primary sources
Go deeper
Use these first-party references to check rules, study the method further, or adapt this guide to your field.