Short answer
Automotive final assembly data covers manual trim, interior, wiring and fastening tasks on a moving line. Existing assets include work instructions, station video and process records; commissioned capture can add viewpoints and labels. None of it contains robot actions, and plant confidentiality, line safety and worker privacy limit what can be recorded or licensed.
Why automotive final assembly draws buyer attention
Final assembly is where a vehicle's body, powertrain and interior come together, and much of it is still done by hand: routing harnesses, seating clips and grommets, installing seats, headliners, door cards and glazing, fitting trim, and driving fasteners to specification. Operations are repetitive, defined by standard work, and performed at stations set by line takt. Buyers evaluating humanoids for this setting ask what data would prepare a robot for it.
The honest answer is that automotive plants hold a lot of material about the work and very little in a form that trains a robot to do it. This page maps what exists, how to judge it, and what must be collected new.
Task families on a final assembly line
| Task family | Examples | Typical challenge for a robot |
|---|---|---|
| Fastening | Driving bolts and screws to torque on trim and chassis parts | Tool reach, torque feedback, access in tight spaces |
| Clip and grommet seating | Door cards, trim panels, seals | Force by feel, deformable parts |
| Harness and connector work | Routing wires, mating connectors | Deformables, hidden contact; see wire harness assembly |
| Large part handling | Seats, cockpit modules, glass | Weight, two-person lifts, moving body |
| Fluids and consumables | Fill, labels, protective films | Tool use, vision of small marks |
| Inspection and rework | Gap and flush checks, repair | Variation and judgment; see rework and repair operations |
The vehicle moves past the station on a conveyor or AGV, so tasks have a moving reference frame. That is a data issue: a fixed camera sees the car pass, and a robot would need tracking or synchronization with the line.
What exists and where
- Standard work and job breakdown sheets. Step lists, times and quality points per station. See standard work and job breakdown sheets.
- Work instructions and training videos. Often by station and model variant.
- Process-monitoring and quality-station video. Fixed cameras recording specific steps for traceability.
- Tool and torque controller logs. Fastening results per bolt and vehicle.
- MES and traceability records. Which vehicle, which station, which time. See MES records aligned to video.
- Tier-one suppliers. Suppliers building modules (seats, cockpits, door assemblies) run similar tasks and may be easier sources than vehicle makers; see contract manufacturers as data suppliers.
What the data can support, and what it cannot
Assembly records and video can support task decomposition, step recognition, cycle-time and variability analysis, visual representation learning of parts and tools in context, and evaluation of step ordering. Fastening logs provide outcome labels for specific joints.
They cannot provide:
- Robot actions or states. Nothing in plant records describes how a humanoid's limbs should move.
- Contact force from the worker. Video does not show how hard a clip was pressed.
- Variant coverage. A plant builds many trims and options; any sample covers some. Held-out variants matter for evaluation.
- Robot-compatible viewpoints. Ceiling or side cameras differ from a humanoid's head and wrist views.
- Performance claims. Public pilot announcements are not evidence of what data works.
Published human-video methods show that human data can help but is combined with robot data; see what human demonstration video can and cannot do.
Safety, confidentiality and permission
Automotive plants may restrict outside recording. Three areas need clear answers.
Line safety. Recording near moving vehicles, powered tools or equipment under service needs the site's safety process. OSHA's lockout/tagout standard requires employers to develop, document and use procedures to control hazardous energy when employees service equipment (OSHA, 29 CFR 1910.147). The standard concerns servicing, not recording, but ask how a collection crew would be briefed on the plant's equipment-control practices and how any camera placement near guarded equipment is approved. See machine safety constraints for recording. This is not legal advice.
Confidentiality. Unreleased models, tooling and process know-how are sensitive. Ask which areas are excluded and how model identity is masked; see manufacturing data confidentiality.
People. Line workers are visible, may be unionized, and may be covered by site policies on recording. Ask about notice, consent and whether faces, badges or names appear. See worker consent and identifiable people. Blurring faces does not by itself settle permission.
Illustrative package: one station, several variants
# Illustrative only - synthetic values, not an available dataset
package: auto_trim_station_v0
station: door_card_install (station_S12), 3 trim variants
video: 2 fixed + 1 head-mounted, constant 30 fps, synchronized
records:
job_breakdown: step_id, step_text, time_allowed_s, quality_point
torque_log: bolt_id, target, result, timestamp
traceability: vehicle_seq_id (hashed), variant_id, station_in/out
labels: step_segments, clip_seated_flag, rework_flag
exclusions: VIN removed, badges masked, pre-release models omitted
rights: training and evaluation use; no redistribution; deletion on request
Limitations and additional preparation
A moving line complicates annotation: the vehicle shifts frame to frame, so object-relative coordinates need tracking. Cycle times are short and workers are fast, so frame rate and shutter speed matter. Variant coverage depends on production schedule during the recording period. Preparing existing material means joining records to video by timestamp, masking identifiers and removing unreleased content. Where the goal is a robot action policy, plan for robot or teleoperated collection in a representative cell.
Choosing a supplier tier
Vehicle makers hold the broadest process records but are the most restrictive on outside recording and the most sensitive about unreleased models. Tier-one module suppliers run similar manual tasks, such as seats, cockpit modules and door assemblies, with a narrower product range and sometimes more flexibility on access. Contract assemblers may offer a middle path. Whatever the tier, ask who owns the process documents: some work instructions belong to the vehicle maker and are provided to the supplier under restrictions.
Also decide how much of the work you need to represent. A single station with several variants gives depth on one task. Several stations give breadth but make annotation and consent more complex.
For the general footage category, see assembly demonstration video.
Questions to ask suppliers
- Which stations and task families are covered, and for which model variants?
- Does the supplier build modules (seats, cockpits) or complete vehicles?
- Are job breakdown sheets, work instructions and torque logs available with timestamps and IDs?
- How is vehicle identity masked or hashed?
- What camera views exist, and are any at hand or head level?
- Which areas or models are excluded for confidentiality?
- What notice and consent applied to line workers, and do labor agreements affect recording?
- How would a new recording be approved for safety, and who must be present?
- Are failures, rework and line stoppages included?
Discuss your requirements
If you are scoping automotive assembly data, describe the stations, variants and permitted uses to SourceX. SourceX looks for US businesses that hold suitable material, checks the data and the supplier's licensing permissions, and coordinates the license and delivery. Nothing is contracted until a supplier agrees, and a request does not guarantee a match. For the wider map, start at the humanoid training data guide or the manufacturing and assembly hub. The site's entry on manufacturing quality records describes related records.
Discuss your humanoid training-data requirements with SourceX
Sources
- 29 CFR 1910.147 The control of hazardous energy (lockout/tagout) — employer duty to document and use energy-control procedures during servicing.