AisleOne

How it fits together

AisleOne, part by part

Scroll or tap a part. Press “See the 3D CAD design” to watch the model, built from our CAD, light each part up and show how it connects.

See the 3D CAD designShow the concept

Real-time 3D render · drag to rotate · final design may varySwipe sideways to rotate

1. 01

   The whole AisleOne

   The aisle's two gondola runs are the aisle's own helper: two cameras in each top canopy watch the shelves opposite and the floor between them, and the brain in run A turns that into restock, spill, label and blocked-aisle calls on the canopy lights and one staff task list.

2. 02Brain

   Raspberry Pi 5 (in run A's top canopy)

   The brain: it runs the aisleone-fw program

3. 

   03

   Run A aisle cameras

   Two cameras in run A's canopy watch run B's shelves (gaps, labels, restock) and the floor from the aisle centre to run B

   Raspberry Pi 5 (in run A's top canopy) · 2 x CSI-2 ribbon · A 12 MP shelf still every 10 min with people blanked; 2 fps floor frames, checked on the brain, never stored

4. 04

   Run B canopy

   The facing run's own Pi 5 and two cameras watch run A's shelves and the other half of the floor, and drive run B's lights

   Raspberry Pi 5 (in run A's top canopy) · Encrypted Wi-Fi link (WPA3, TLS 1.3) across the aisle · Run B's gap, label, spill and traffic results; light and tone commands back

5. 05

   Call lights and tone

   Amber over a shelf gap, red with a tone over a spill or a person down, in both canopies

   Raspberry Pi 5 (in run A's top canopy) · One data wire + GPIO · Call colour, position along the run, tone

6. 06

   Secure element and tamper switch

   Signs every call; marks the log and turns the cameras off if a canopy cover is opened

   Raspberry Pi 5 (in run A's top canopy) · I²C + GPIO · Signatures, tamper flag

7. 

   07

   Power from the canopy light feed

   Turns the canopy's own lighting feed into 5 V inside the closed canopy

   Raspberry Pi 5 (in run A's top canopy) · 5 V rail · Power, about 6.5 W per run (est.)

8. 08

   All connected

   6 key parts, one program, and your phone

Our circuit board · rev A-0

One board of our own inside AisleOne

Every chip on one board we design, shaped to fit: the processor, cameras and sensors, Wi-Fi and Bluetooth, lights, sound and power. No plugs, ribbons or adapters. It is checked in our CAD against every part around it, and it arrives from the maker ready to run our software from a microSD card. Turn the model, or pick a part below to see where it sits.

Demo · 6 real-world moments

See AisleOne work

Short animated scenes of real moments: what happens today, what AisleOne does about it, and what lands on your phone. The same scenes play in the IEMI app.

Demo scenes show how AisleOne and its app are designed to behave. AisleOne is in design; nothing here is recorded from real equipment yet.

The idea

The aisle's own shelves spot the gaps, call spills, check the restock and give the store one task list

How the parts connect

AisleOne, wired up

Every part, what links it to the brain, and how the data reaches your phone. A solid line is a cable; a dashed line is a radio link.

The program

What controls AisleOne

while True:
    for cam in cams: floor = floor_change(cam.frame(lowres=True), clean[cam])
    if tick(600) and no_shopper(): gaps, labels = score_still(cam.still(), prices)
    calls = spill_rule(floor) + sign_rule(floor) + gap_rule(gaps) + label_rule(labels)
    calls += blocked_rule(objects) + down_rule(pose) + runb.calls()
    for c in calls:
        if c.conf < c.line: c.kind = 'go and look'   # never guess
        atecc.sign(c); canopy(c.run).show(c); staff.task(c)
    time.sleep(0.5)

To your phone

How the data reaches AisleOne staff app

Staff phones and the manager's tablet (Android and iPhone)

1. Gondola run A
2. Store Wi-Fi (WPA3) · MQTT over TLS
3. IEMI cloud (the store's account)
4. Push
5. Staff phones and the manager

Key specs

Targets for production units; they may change during development.

Pricing

Target pricing (stage 2, 29 Oct 2026; was $690). Bought separately per aisle: two ShelfScan cameras ($598 + $58/month), a share of a SpillGuard system (about $100 + $33/month) and a plain lit canopy pair (about $240): about $938 + $91/month (est.). Parts about $704 per aisle at qty 1.

Indicative launch pricing in USD, excluding taxes and shipping. Final pricing is confirmed before any balance is charged.

1. 4 Nov 2026, 06:30LatestA grocer buys AisleOne as the aisle's own pair of gondola runs, fitted by the shelving installer and left in the store for years. So the electronics must be boards an assembly house makes and tests the same way every time, inside each run's 3,660 × 460 × 80 mm top canopy, on the canopy's own light feed, never below 1.83 m and never past the base deck into the aisle (ADA 307.2).

   • What we designed: boards of our own in the top canopy of each gondola run, the same set in both. An 8-layer main board lies on the canopy floor at the middle of the run with the processor and a heat block into the aluminium canopy, memory, the security chip, the latched microSD card and a small piezo over the tone holes. Two camera chips sit on small tabs off its front edge, each turned 30° out to its end of the aisle and 30° down, the lens just inside the cover. A light strip stands behind the opal cover with 48 call lights, the Wi-Fi module and a magnet sensor for the cover; two small light boards at the run's ends carry 12 more each; a power board at the back takes the canopy's own light feed through a fuse to a soldered supply.
   • How we checked it: the boards are written as code and opened in KiCad (0 design-rule violations). The code places 110 parts at their sizes and checks no overlaps, the heights in every gap, the lights behind the opal, the piezo over its holes, the gap between the mains and 5 V sides, and that all 28 aisle calls from the CAD (the opposite bottom and top shelves at the end and the middle, the aisle floor at the end and at the opposite base deck, a person down) are in view from where each lens really sits. The shared design review: 15 pass, 2 fixed, 11 to prove, 0 fail, with a request for quote written from the parts.
   • What we learned: the bench cameras sat a quarter of the way along each run on 1 m ribbons, too far for a camera chip on our board, so both moved to the middle and turned out; the CAD's windows and tone holes moved with them and every view still passes. From the middle the farthest price label is 2.7 m away, and the review showed an 8 MP chip under a wide lens gives its digits 6.4 px where the reader needs 8; a 12 MP chip under a narrower lens covers every call and gives 8.6 px, which we prove on the first boards. There were no new CricketOne board decisions since the last update, and every other board was re-run and still passes.
   • What is next: the camera brackets and the boards inside the canopy CAD with every part checked in 3D; Wi-Fi range from inside the aluminium canopy across the aisle; the label read on real shelf labels; and the next board, CourtOne.

   See it in 3DFull research and design

2. 29 Oct 2026, 23:41US weights and measures inspectors check shelf prices against the till with the NIST Handbook 130 price verification procedure, where a store passes when at least 98 % of sampled prices match (NIST). In the EU, AI that reads the emotions of people at work has been banned since February 2025 (AI Act Article 5), and a store aisle is a workplace for the staff who restock it.

   • Market and rules check (DESIGN.md): the NIST Handbook 130 price verification procedure (2026 edition), the EU AI Act ban on emotion reading at work, Walmart's store cameras now calling low shelves and spills (October 2026 report), a regional grocer's hourly shelf-camera pilot and shelf vision spreading to chains in Asia, each with what it changes in AisleOne (sources R1-R7)
   • New assist call: a weekly price-accuracy check gives each aisle's label match rate against the 98 % guide and the labels to fix; below 85 % of labels read it gives only the list. No new part and no CAD change, now 9 calls
   • Privacy: staff are masked like shoppers, restock times are kept per aisle and bay and never per person, no staff scores and no emotion reading; two new rule lines in equipment.json
   • Market: the largest chain now wants gaps and spills from one camera system, but shelf cameras are still sold to chains on contract; price kept at $920 per aisle + $15/mo
   • Product page https://iemilabs.com/products/aisleone/ gains a price-check FAQ, a staff privacy FAQ and an updated privacy spec
   • Next: ask independent grocers in the phase 3 buyer review whether a weekly price-accuracy figure helps them prepare for an inspection; close-out C9 (recommend the first product to the owner) from 30 Oct 00:00

   See it in 3DFull research and design

3. 29 Oct 2026, 13:55Before customer reviews start in phase 3, each of our eleven industry products needs a list of who to ask, its three open questions and a short survey. This is close-out task C8 in our phase 1 plan.

   • Section "Industries" in products/BUYER-REVIEW.md: AisleOne, RoomOne, DockOne, CellOne, ChillOne, CareOne, FarmOne, PumpOne, SiteOne, BayOne and RoadOne each get who to ask (the payer, the people on the shift, the channel that sells it and the person who accepts the record, with sources for what they pay today), the three open questions from their stage 2 refinement, a five-question survey and what each answer would change
   • Extra rules for industry reviews: we name kinds of business, never firms or people; we meet in an office or by call, and on a site walk we follow the site induction and take no photos of workers, patients, residents, guests or the public; no question suggests a product makes a place safe or compliant
   • What we learned: ten of the eleven ask whether buyers would replace a working item or want a kit for the one they own, every product asks which call pays for it first, and in regulated places whether an auditor, inspector or insurer accepts the signed record decides the purchase
   • Next: close-out C9 from 30 Oct, scoring all 24 products to recommend a first product to the owner

   See it in 3DFull research and design

4. 29 Oct 2026, 00:26An independent grocer needs shelf gaps found and refilled, spills called and the wet-floor sign checked while two staff run the store: independents ran 6.6 % out of stock in fiscal 2025 (FMS/NGA via The Shelby Report), and the shelf cameras and robots that do this are sold to chains on contracts with no public price (Vusion, Simbe, Focal Systems, Oct 2026).

   • Buyer check with sources: what stores pay for shelving (about $150-350 per 4 ft section), and the closest products: Vusion Captana rail cameras (Carrefour, Feb 2026), Simbe Tally 4.0 robots (regional grocers since June 2026) and icetana spill alerts on existing CCTV, all priced on contract
   • What we learned: shelf vision is moving from Walmart down to regional grocers, but nobody sells it as the shelving itself, and nobody covers shelf gaps and spills in one item
   • Brains changed: the Raspberry Pi 5 4 GB now costs $130 (was $60), so both canopies use the 2 GB model at $90 on the same board outline; the CAD fit and all 28 camera views are unchanged
   • Parts list prices checked on shop pages today: $704 of parts per aisle at qty 1; cost and margin table in the design notes (about 13 % at qty 1 and 36 % at qty 100, est.)
   • Price raised from $690 to $920 + $15 a month per aisle, still under about $938 + $91 a month for the separate items (about $1,100 against $2,030 in the first year)
   • Product page: the first three FAQs answer the buyer's top questions (why not shelf cameras or a robot, does it recognise faces, what it costs), plus power and size specs
   • Next: three questions for grocers in customer reviews (price per aisle or per store, retrofit or from the gondola maker, which call pays first), then RoomOne for facilities and hospitality

   See it in 3DFull research and design

5. 28 Oct 2026, 08:24An independent grocer needs one thing in each aisle that finds shelf gaps and checks the restock, calls spills and the wet-floor sign, and flags wrong labels. Independents ran 6.6 % out of stock in fiscal 2025 (FMS/NGA via The Shelby Report), and shelf cameras are sold to chains with no public price (Supermarket News).

   • New CAD: two 3,660 mm gondola runs with 460 x 80 mm top canopies and printed opal covers; in each canopy a Raspberry Pi 5, two Camera Module 3 Wide aisle cameras, four call lights, a tone, secure element, tamper switch and a potted power supply; 22/22 parts fit from datasheets, 28 camera views checked, 1,830 mm clear aisle and the canopy behind the base deck (ADA), no collisions, no floating parts
   • equipment.json: ADA 403.5.1 and 307.2, UL 962, FTC Rite Aid order (no face recognition) and ICO rules; 8 calls (6 alone), 3 impact cases and every security key
   • idea.json aisle plan, integration.json (5 parts, 5 app screens with a 'no staff in the aisle' mode), bom.csv (about $631 an aisle), buy.json with 6 key parts
   • New product page at /products/aisleone/ in Retail: $690 + $15/month per aisle against about $938 + $91/month bought separately (est.); folds StoreEye, ShelfScan and SpillGuard; next: F8 Facilities and hospitality

   See it in 3DFull research and design

6. 28 Oct 2026, 08:24An independent grocer needs one thing in each aisle that finds shelf gaps and checks the restock, calls spills and the wet-floor sign, and flags wrong labels. Independents ran 6.6 % out of stock in fiscal 2025 (FMS/NGA via The Shelby Report), and shelf cameras are sold to chains with no public price (Supermarket News).

   • Now part of AisleOne: https://iemilabs.com/products/aisleone/

   See it in 3DFull research and design

7. 28 Oct 2026, 08:24An independent grocer needs one thing in each aisle that finds shelf gaps and checks the restock, calls spills and the wet-floor sign, and flags wrong labels. Independents ran 6.6 % out of stock in fiscal 2025 (FMS/NGA via The Shelby Report), and shelf cameras are sold to chains with no public price (Supermarket News).

   • Now part of AisleOne: https://iemilabs.com/products/aisleone/

   See it in 3DFull research and design

8. 28 Oct 2026, 08:24An independent grocer needs one thing in each aisle that finds shelf gaps and checks the restock, calls spills and the wet-floor sign, and flags wrong labels. Independents ran 6.6 % out of stock in fiscal 2025 (FMS/NGA via The Shelby Report), and shelf cameras are sold to chains with no public price (Supermarket News).

   • Now part of AisleOne: https://iemilabs.com/products/aisleone/

   See it in 3DFull research and design

Smart equipment · legal on the field

AisleOne is the store gondola shelving, made smart

It replaces the aisle's existing two gondola runs (or their top canopies on the same uprights), at the same place on the store floor, with every part built in. Everything inside the two top canopies: in each, a Raspberry Pi 5, two aisle cameras, four call lights, a tone, a secure element, a tamper switch and a potted supply. About +1.1 kg per run over a plain lit canopy (est.); no ceiling camera, shelf-edge camera, floor robot or wall box.