Multifunction steering wheel control with AI — five worlds in a single part, and no cross-check in time.

The problem

Five different worlds, five quality systems, one single steering wheel.

The modern multifunction steering wheel is the most complex part of the interior — and every subsystem lives in its own quality system:

1. The leather or synthetic on the rim — stitched by hand or by machine, a critical cosmetic defect that the OEM rejects on appearance.
2. The foam over the metal armature — hardness, uniformity, color. One bad mix and you feel it in the first month.
3. The driver airbag module — bolted to the hub with critical torque and angle, with its own IATF and UNECE standards.
4. The physical and capacitive controls — each with its own electrical test per model and trim level. The CAN/LIN test validates that the cruise control button appears as such in the ECU.
5. The CAN/LIN connector to the ECU — specific pinout, identification frame, validation against the test bench.

The quality manager knows it. The operator has their test bench, their smart nutrunner and their model card. The classic system works when the line is running at steady state — and the defects concentrate at the multi-model SKU changeover, exactly when a Sport steering wheel in gray leather gets mixed in with a Comfort in beige leather and nobody notices until the OEM rejects it at receiving.

How it fits the IRIS system

iLEAN does not replace the test bench — it puts the five worlds in the same sentence.

The multifunction steering wheel problem is not a lack of benches: it is information in islands (electrical test bench result, airbag module torque, image of the leather, foam hardness, model recipe). At the critical moment — the steering wheel leaving the line — those five sources are not cross-referenced. iLEAN acts as the putty that fills those gaps without touching the bench or the nutrunner.

Edge sees the leather, the controls and the mounted airbag. Connect reads the test bench, the airbag torque and the model recipe. The agent cross-references and holds. The person signs — never the other way round.

The three iLEAN pieces applied to multifunction steering wheel control:

• Edge — a machine vision terminal (CNN) over the critical stations: image of the stitched leather (stains, loose threads, color), of the airbag module aligned to the hub, of the controls on the rim (play, marking, trim cover). It triggers the actuator (light stack, diverter) in milliseconds if something does not add up. It works with no network: if the plant loses WiFi, Edge keeps reading and holding units.
• Connect — captures the CAN/LIN frames from the electrical test bench (which button responded, which frame was emitted), the torque and angle from the airbag module nutrunner, the model recipe from the ERP/MES and the module serial number from the laser marking. What arrives from outside (an email from the leather supplier about a batch under suspicion for color) enters at second zero.
• Agent — cross-references the image of the leather, the test bench result, the airbag module torque, the steering wheel serial number, the destination VIN and the model recipe. If something does not add up — a Sport steering wheel with beige leather from an SKU error, a button with no CAN response, a torque outside the window — it holds the unit and alerts the quality manager. The person validates and signs; the line does not restart on its own.

See the full IRIS architecture →

Before and after

Test bench + human eye vs. in-line control with iLEAN

Aspect	Today (bench + checklist + human eye)	With iLEAN Edge + Connect + Agent
Cosmetic leather defects	End-of-shift visual inspection, subjective	CNN vision on the stitched rim, before the trim cover
Electrical test of the controls	Result on the bench PC, not linked	CAN/LIN frames captured and linked to the serial number
Airbag module torque	Recorded in the nutrunner, not linked to the VIN	Captured by Connect and linked to the destination VIN
Multi-model SKU changeover	Paper card, risk of mixing during the transition	Edge validates color/trim against the model recipe
Batch-to-VIN link	Rebuild it if there is a PRR, weeks	File per unit, automatic, ready for PPAP/8D
Operation with no network	n/a	Edge keeps running on the light from the panel

Impact estimate

Impact estimate for your plant — to be validated with your numbers.

The block below is an estimate to be validated with the specific data of your plant. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

• Multi-model multifunction steering wheel assembly line, with a leather sewing machine, an airbag module nutrunner and a CAN/LIN electrical test bench at the end.
• Edge pilot on one line (camera over the leather + camera over the controls + integration with the test bench and the nutrunner). First value expected within a few weeks.
• Indicative payback between 4 and 9 months, depending on the frequency of OEM PRRs and of units blocked at the OEM's plant over the past years.
• The hard lever is a single sorting campaign avoided at the OEM from a multi-model SKU mix-up, plus the aggregate cost of cosmetic PRRs.

And the quality manager's reasonable doubt

“What if the AI confuses gray leather with beige when the booth lighting changes?” — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI simply compares a color measured in a calibrated booth with a model reference, the best models brought error below 1.5% ^[1]. And even so, what is critical is never decided alone: iLEAN holds the unit and the person signs. The automotive quality standard on the order of 25 PPM ^[2] cannot be held with sampling: it is held with vision on every part.

^[1] OpenAI paper “Why Language Models Hallucinate”, 2025 — on the reliability of AI in anchored tasks.

^[2] Symestic — automotive quality standard on the order of 25 PPM.