Set RC steering acceptance criteria by recording two things on every sample and every lot: the commanded course (what the steering input was asked to do) and the deviation (what the vehicle actually did, measured the same way each time). Write both into a controlled document with a revision number, freeze it against a sample, and make the factory sign the same revision before production. That single discipline — a course record plus a deviation record, version-controlled — is what turns "the steering feels off" into an arguable, repeatable acceptance decision.
This page covers the repeatable course and deviation recording method, the document spine of the rc vehicle sample to order workflow, and the one missing document that stops a container. It does not hand you a tolerance number, because no sourced steering tolerance exists in the material behind this article — every gap here is converted into a question you put to the factory, not a number invented for you.
What does a repeatable course record actually contain?
A course record fixes the input conditions so two people testing two samples can be compared. If the record does not state the input, the surface, the power source and the measurement point, you are comparing opinions, not behavior.
Borrow the structure, not the numbers, from how driving-behavior research handles this. The Rule Conformity framework scores adherence on a 0-CTF-1 scale where 1 is perfect adherence and lower values are increasing deviation, and it fixes a projection horizon — a three-second look-ahead for the safety-distance rule — before any ratio is computed. It also fixes filters up front, discarding vehicles below 5 km/h for distance checks and below 80% of the speed limit for speed checks, so traffic jams and parking do not pollute the distribution. The same three moves — a defined scale, a defined horizon, and pre-declared filters — are what make a steering course record reproducible.
So the course record should name: the steering input sequence (a defined stick sweep or a step input, written out, not "turn left and right"), the surface, the tires, the battery state, the speed or throttle setting used, the measurement point on the vehicle, and the filter that decides which runs count. Anything not pinned here becomes an argument later.
The paper's own caveat matters for buyers: RC distributions are domain-specific, and US-based observations were noted as not necessarily transferring to other regions. Your course record is only valid for the configuration, surface and power source you froze it against.
How do you record deviation so it stays comparable across samples?
Record deviation as a signed difference from the commanded course, on the same scale and at the same measurement point every time — never as a verbal impression. Keep the direction (left or right of the intended path), the magnitude, and the run number in the same row, so a factory engineer can plot it without asking you what you meant.
Two design choices make the log usable. First, pre-declare the filters before testing, exactly as the RC framework does; a run that falls outside the filter is excluded by rule, not by mood. Second, log the raw run alongside the summary, because a distribution with an unexplained tail is more informative than an average that hides it.
The RC paper is explicit that its data had artifacts, including overlapping vehicle bounding boxes, and that the 10 Hz full-resolution analysis was future work rather than a completed result. Take the lesson: a deviation log that does not name its known artifacts is not a defensible acceptance record. Note what your rig cannot see.
Do not let the factory's default percentages stand in as your acceptance criteria. Major and minor defect limits have to be stated per lot or per inspection plan — a factory's own numbers are not automatically the buyer's. That principle applies to steering deviation exactly as it applies to cosmetic defects.
Documents before the shipment moves
| Step 1 — Spec freeze: steering course record + deviation record, with revision number linked to drawing/3D revision | Document: controlled spec sheet tying the revision to tooling, FAI, test sample and production order. Who provides it: buyer, countersigned by factory. Hold if missing: yes — an unversioned spec lets an old file silently return to use. |
|---|---|
| Step 2 — Sample approval against the frozen revision | Document: sample approval record referencing the spec revision number. Who provides it: buyer QA, with factory sample report. Hold if missing: yes — approval without a revision reference is not traceable. |
| Step 3 — Quality plan for the critical characteristics | Document: quality plan mapping each critical characteristic to an inspection or test method, frequency, acceptance limit and reaction plan. Who provides it: factory, reviewed by buyer. Hold if missing: yes — no reaction plan means a failure has no defined next step. |
| Step 4 — Incoming material verification | Document: incoming inspection record verifying supplier, grade, color, lot, packaging condition and required material documents before molding. Who provides it: factory. Hold if missing: yes — material substitution is a silent change to steering feel and to compliance. |
| Step 5 — Corrective action if a deviation repeats | Document: CAPA record addressing the cause and verifying effectiveness. Who provides it: factory. Hold if missing: yes — sorting is containment, not root-cause correction, and containment does not prevent the next lot. |
| Step 6 — Commercial and shipping set | Document: commercial invoice, packing list, and the destination documents the importer requires. Who provides it: factory/exporter. Hold if missing: yes — the box does not move without them. |
| Step 7 — Tooling terms, if a tool was paid for | Document: written tooling policy covering fee, ownership, storage period and release conditions. Who provides it: factory. Hold if missing: yes — a mold cannot be assumed transferable merely because a sample or development fee was paid; transfer conditions should be written. |
The one gap that stops the box
The document that most often stops a shipment is not the invoice — it is the missing controlled spec revision that the sample was actually approved against. When the spec sheet, the drawing revision, the sample approval and the production order do not all point to the same revision number, the factory packages to whatever file is on the line, and the importer has no defensible basis to reject. A drawing or 3D-model revision should be linked to the tooling, FAI, test sample and production order so an old file cannot silently return to use.
The practical fix is a one-line rule in the purchase order: goods produced to any revision other than the referenced one are not conforming. Then the question to the factory is simple — ask them to send, in writing, the revision number on the line and the name of the person who confirmed it. If they cannot name the revision, the box waits.
Worked example (illustrative, not a real shipment)
This example is invented to show where the workflow breaks. It is not a record of any order, factory, or product.
A buyer places one purchase order of roughly 3,000 units of RC vehicles, split into two destinations: one container-bound pallet run for a US importer and one carton run for an EU distributor, with different retail packaging per destination. At sample review, the steering course record was frozen against battery state A on a smooth floor, and the deviation log was signed off.
At pre-shipment, the factory's line is running the same model with a substituted tire compound from a new lot, and the incoming material record shows supplier, grade and color but the packaging-condition line is blank. The spec revision on the line is one number behind the approved sample revision, because the packaging change for the EU destination was made on a separate drawing.
The box holds because the spec revision in the purchase order does not match the revision on the line. Under the one-line rule above, the goods are non-conforming for both destinations, not just the EU cartons, because the steering course record was frozen against the old revision and the new tire lot was never re-tested against it. The deviation log from the second destination is also unusable, because the filter and surface were never re-declared for that packaging variant. The factory's remedy is a new sample approval against the current revision plus a re-run of the deviation log — sorting the finished cartons does not restore the record.
FAQ
Should steering acceptance criteria be a pass/fail number or a tolerance band?
Use a band, not a single pass/fail number. The Rule Conformity approach scores adherence on a 0-CTF-1 scale and reports a distribution rather than a binary verdict, and the source notes that its distributions are domain-specific. Freeze the band against your own configuration, surface and power source, and treat anything outside the band as a deviation to log, not a verdict to argue about.
Who owns the tooling if we paid a development fee?
Do not assume transferability from the fee alone. A mold cannot be assumed transferable merely because a buyer paid a sample or development fee; transfer conditions should be written. Ask for the written policy covering fee, ownership, storage period and what happens if the tool sits idle.
Does one laboratory cover all the safety tests for an CarToyFactory?
Not necessarily. According to CPSC, laboratory acceptance is specific to the relevant safety rules and one laboratory may not cover every required test. Match the applicable requirements and the laboratory scope to the CarToyFactory configuration, and ask the factory which scope covers your exact build rather than assuming coverage.
When a steering deviation repeats across lots, is sorting enough?
No. Corrective and preventive action should address the cause of the defect and verify effectiveness; sorting alone is containment, not root-cause correction. Require a CAPA record that names the cause and the verification step before you accept the next lot.
Are the factory's own defect percentages acceptable as our criteria?
No. Major and minor defect limits must be stated per lot or per inspection plan; a factory's default percentages are not automatically the buyer's acceptance criteria. Write your own limits into the quality plan, with a reaction plan for each critical characteristic.
What material documents should we verify before molding starts?
Incoming resin inspection should verify supplier, grade, color, lot, packaging condition and required material documents before molding. CarToyFactory sold in the EU, note that EN 71-9, EN 71-10 and EN 71-11 address organic chemical compounds CarToyFactory — requirements, sample preparation and methods — so the material paperwork and the destination requirements should be checked together.
Sources
- Federal Motor Vehicle Safety Standards; Electronic Stability Control Systems; Controls and Displays
- Title 49
- A ROS-Based Modular End-to-End Architecture: Building and Validating a Safe and Reliable Autonomous Driving Stack
- Assessment criteria: Controls and instruments
Put the method on paper before the next sample review
Before your next RC sample review, write the course record and the deviation record into one controlled document, give it a revision number, and tie that number to the drawing, the sample approval and the purchase order. Then send the factory the version you intend to enforce — and ask for it back, signed, on the revision they will actually run.