A Wi-Fi locator discarded up to 95% of router signals—inside one building
A peer-reviewed study kept 8–26 of 172 Wi-Fi access points with less than a 0.6-percentage-point accuracy change on one public benchmark. The catch: most zero-shot context transfers failed, and battery savings were estimated rather than field-measured.
The short version
3 key points
- What changed
- On JUIndoorLoc—25,364 fingerprints, 172 access-point features, three floors and four Android devices—the method retained 8–26 access points per floor.
- Why it matters
- A proven reduction in radio scanning could make indoor guidance lighter on batteries and cheaper to run on ordinary phones, but this study did not measure battery life or whether people reached destinations more reliably.
- What to watch
- Independent multi-building replication, direct phone battery measurements, stability after router changes, comparisons with alternative selection methods and end-to-end navigation tests with users.
Most router signals were expendable on this benchmark
Researchers from Mansoura University and Egypt's Military Technical College tested an entropy-based way to rank Wi-Fi access points for indoor positioning. They used JUIndoorLoc, a public dataset collected in one Jadavpur University building in India: 25,364 signal fingerprints, 172 access-point columns, three floors, four Android devices and observations spread across 31 days and several contexts. A Random Forest was trained with stratified five-fold cross-validation, with feature thresholds recalculated inside each training fold to limit information leakage. The selected sets kept 17 access points on floor 3, eight on floor 4 and 26 on floor 5—reductions of 90.11%, 95.34% and 84.88%. Compared with using all 172 signals, the reported accuracy changes were -0.41, -0.51 and -0.42 percentage points. A K-nearest-neighbours check lost 2.5, 0.5 and 0.4 points, suggesting the result was not confined to one classifier, although floor 3 was more sensitive.[1]
Why less scanning could matter to people
Indoor navigation systems can use Wi-Fi fingerprints where satellite positioning is unreliable, including campuses, hospitals and warehouses. Fewer access points mean smaller models and less work per location estimate: the paper reports a Random Forest model shrinking from 3.82 MB to 0.24–0.73 MB and inference time falling from 1.45 milliseconds to below 0.27 milliseconds. Under simulated signal noise, the compact floor-5 set remained more accurate than all features at each tested noise level. The authors also estimate that scanning only selected channels could cut radio-on time from roughly 300 to 60 milliseconds, implying about 80% lower scanning energy. That last figure is an engineering estimate, not a measured battery result, and the study did not test navigation success, accessibility or safety with real users.[1]
Who it affects
A proven reduction in radio scanning could make indoor guidance lighter on batteries and cheaper to run on ordinary phones, but this study did not measure battery life or whether people reached destinations more reliably.[1]
Keep in mind
One public dataset from one Indian university building; floor classification rather than full navigation; simulated noise and estimated energy use; weak zero-shot transfer between collection contexts; no external-building or user trial.
Source analysis · 1 linked source. Original source date: 18 October 2026.
Read the full report · 4 min The evidence, methods, implications and limitations
Most router signals were expendable on this benchmark
Researchers from Mansoura University and Egypt's Military Technical College tested an entropy-based way to rank Wi-Fi access points for indoor positioning. They used JUIndoorLoc, a public dataset collected in one Jadavpur University building in India: 25,364 signal fingerprints, 172 access-point columns, three floors, four Android devices and observations spread across 31 days and several contexts. A Random Forest was trained with stratified five-fold cross-validation, with feature thresholds recalculated inside each training fold to limit information leakage. The selected sets kept 17 access points on floor 3, eight on floor 4 and 26 on floor 5—reductions of 90.11%, 95.34% and 84.88%. Compared with using all 172 signals, the reported accuracy changes were -0.41, -0.51 and -0.42 percentage points. A K-nearest-neighbours check lost 2.5, 0.5 and 0.4 points, suggesting the result was not confined to one classifier, although floor 3 was more sensitive.[1]
Why less scanning could matter to people
Indoor navigation systems can use Wi-Fi fingerprints where satellite positioning is unreliable, including campuses, hospitals and warehouses. Fewer access points mean smaller models and less work per location estimate: the paper reports a Random Forest model shrinking from 3.82 MB to 0.24–0.73 MB and inference time falling from 1.45 milliseconds to below 0.27 milliseconds. Under simulated signal noise, the compact floor-5 set remained more accurate than all features at each tested noise level. The authors also estimate that scanning only selected channels could cut radio-on time from roughly 300 to 60 milliseconds, implying about 80% lower scanning energy. That last figure is an engineering estimate, not a measured battery result, and the study did not test navigation success, accessibility or safety with real users.[1]
The distinction that matters
How the compact locator was tested
The experiment reduced the signal inputs first, then compared accuracy and robustness with the full 172-feature benchmark.
Collect
Use 25,364 fingerprints from one building
172 Wi-Fi signal features across three floors
Rank
Score each access point by entropy
Retain 8–26 signals per floor
Stress-test
Compare cross-validation, noise and dropout
Small same-building loss; weak zero-shot transfer
One building is the hard limit
The strongest warning comes from the paper's own transfer test. Training in one collection context and applying the model unchanged to another generally produced near-zero accuracy, showing that device, time and environmental changes can overwhelm a compact fingerprint. Adding five labelled target examples per class restored reported accuracy to 0.76–0.99, but those tests still came from the same building and dataset. The study therefore supports a narrower claim: entropy ranking can remove many weak or redundant signals while preserving same-dataset floor classification, and a small amount of local calibration may help when conditions shift. It does not show that eight access points will work in another hospital, campus or warehouse, nor that the selection remains stable after routers move or disappear. Confidence would rise through preregistered tests in multiple buildings and countries, comparisons with other feature-selection methods, longitudinal router changes, and phone-level trials measuring battery draw and end-to-end navigation errors. The article is peer reviewed and open access; publication charges were funded by Egyptian Knowledge Bank, and the authors declared no competing interests.[1]
Watch the background, then check the evidence
Illustrative photo · not a video frame
The Royal Society · 2 min 31 sec · Background explainer
What is artificial intelligence? | The Royal Society
YouTube loads only when you choose; playback starts when you press play.
A short introduction to artificial intelligence from the Royal Society. Start here for basic concepts, then use current reporting to explore what has changed.
Evidence check: Published in 2018, before today’s generative AI tools. Useful for basic concepts, rather than an overview of current capabilities.
Published 2018-11-28. Source checked 14 October 2026. Check captions on the publisher’s video page.
Explore current AI technologyPhoto: ThisisEngineering / Unsplash · Licence
Evidence check
One public dataset from one Indian university building; floor classification rather than full navigation; simulated noise and estimated energy use; weak zero-shot transfer between collection contexts; no external-building or user trial.
Reporting basis: Source analysis. Original source date: 18 October 2026.
Continue the story
Read next
Also about edge computing
Cloud offloading could change the performance and battery trade-off for robots
Microsoft Research reports that moving some physical-AI inference from a robot's onboard processor to edge or cloud GPUs can improve task success, efficiency and the complexity of workloads the machine can handle.
Read the 2 min reportAlso about edge computing
Can an AI explanation survive model compression?
A study of three image models found that INT8 conversion could preserve predictions while moving the pixels highlighted as important. A poor target-layer choice also created an apparent explanation collapse that disappeared when corrected.
Read the 6 min reportExplore the full source and evidence record
What this means for people
- A proven reduction in radio scanning could make indoor guidance lighter on batteries and cheaper to run on ordinary phones, but this study did not measure battery life or whether people reached destinations more reliably.
Global context
The authors are based in Egypt and evaluated an Indian benchmark, but the evidence covers one building. Wi-Fi layouts, device mixes, maintenance practices and accessibility needs vary widely across regions and settings.
What the evidence does not yet show
- One public dataset from one Indian university building; floor classification rather than full navigation; simulated noise and estimated energy use; weak zero-shot transfer between collection contexts; no external-building or user trial.
What to watch next
- Independent multi-building replication, direct phone battery measurements, stability after router changes, comparisons with alternative selection methods and end-to-end navigation tests with users.
Living evidence record
Impact record IAI-1TD3RI2
Evidence stage
Studied
Confidence
Supported
Reporting basis
Source analysis
Independent or research support
Present
Record status
Monitoring
Last checked
18 October 2026
Source trail
1 direct source across 1 source type.
People impact
Documented in this record.
Uncertainty
Limits and next checks are explicit.
Stages describe the evidence available—not whether a technology is good or bad. See the public method.
Single-source reporting disclosure
This record analyses one direct source. It can establish what Scientific Reports published or reported, but it is not independent corroboration of every performance claim or predicted outcome. The confidence label will change only when broader evidence is added.
Evidence trail
Sources used for this report
Links checked 18 October 2026
Research question, topics and regions
Whether selecting a small, information-rich subset of Wi-Fi access points can preserve indoor floor-classification accuracy while reducing computation and radio scanning
Topics: Indoor navigation · Wi-Fi · Edge computing · Battery use · Domain shift
Regions: India · Egypt · Asia · Middle East
This report is labelled source analysis. We summarise and analyse source material in our own words; company statements remain attributed claims until independently supported. Translated summaries preserve the meaning of the original source and link back to it. Read our editorial standards.
The Impact Brief · Free
Keep following the stories that matter to you.
Get our weekday briefing on what changed and why it matters. Choose your topics below.
Reader commentary
What does this mean where you are?
Tell us what you have seen at work, in your community or in everyday life. A question is welcome too. No account is needed; we review notes before publishing them.
Explore commentary across the portal →Published reader notes
0No published reader notes yet. You can start the evidence-led discussion above.
Prefer a private correction or response? Contact the newsroom.