Short answer
Automotive service is attractive for robot data because tasks such as oil changes, tire rotation, brake inspection and battery replacement are standardized and heavily documented. Dealer service departments and repair chains hold repair orders and parts records; new video can be recorded in their bays. Records give task frequency and parts; video gives visual process; neither provides robot actions or forces.
Why vehicle service stands out
Few maintenance domains are as regular. A service bay performs the same operation on the same model many times, against a published procedure, with a flat-rate time and a coded repair order. Vehicles are also instrumented with standard interfaces and have a clear parts ecosystem. The combination makes it a candidate for studying how task data transfers across variants: same task, different vehicle.
The domain also has hard physical demands. Wheel removal involves heavy loads and lug nuts at specified torque. Underbody work happens beneath a lifted vehicle, in poor light, with dirt and fluids. A humanoid's reach, payload and balance matter as much as its learning. This page addresses the data side; the maintenance hub and the humanoid data sourcing guide place it in context, and bicycle and light vehicle repair covers smaller-scale variants.
Task anatomy for common services
| Service | Core steps | Variation that matters |
|---|---|---|
| Oil and filter change | Lift, locate drain plug, drain, replace filter and gasket, refill, check level | Filter location, drain plug type, capacity |
| Tire rotation | Lift, loosen lugs, remove wheels, reposition per pattern, torque in sequence | Wheel type, locking nuts, rotation pattern |
| Brake inspection | Remove wheel, inspect pad thickness, rotor, caliper, hoses | Caliper design, wear sensors |
| Battery replacement | Disconnect terminals in order, unclamp, lift, install, reconnect, check | Location, hold-down type, registration needs |
| Cabin and engine air filter | Open housing, remove, orient, close | Access panel design |
| Wiper and lamp replacement | Release, remove, fit new, test | Connector type |
Failure modes include cross-threaded drain plugs, over-tightened lugs, missed clips on housings, wrong filter, a dropped part under the vehicle and a tool slipping on a rounded fastener. Each shows up differently in video: some visibly, some only in a later comeback record.
Two kinds of data, two kinds of structure
Records. A repair order lists the vehicle (year, make, model), the complaint, the work performed, parts, labor operation codes and time. Fitment and part data in the aftermarket follow Auto Care Association standards: ACES is the standard for managing and communicating product fitment data for light, medium and heavy duty vehicles, powersports and off-highway equipment, and PIES covers product information such as attributes, descriptions, images and packaging. Both are machine-readable XML and rely on subscription reference databases (Auto Care Association). A buyer who wants part identities that join cleanly to repair orders should ask whether suppliers use these standards. See spare parts identification for the part-recognition side.
Repair-order narrative text is terse. MaintNet includes automotive logbook data together with abbreviation lists and pre-processing tools (Akhbardeh et al., 2020), which illustrates the normalization effort for "R&R FR BRK PADS" and similar shorthand. Work order histories discusses how to read these records for symptom, fix and parts.
Video. Dealer training libraries, manufacturer technician-training content and shop cameras exist, but footage of real service is not routine, and unions, privacy and liability concerns limit it. New recordings with head-mounted cameras and a fixed bay camera can be arranged with the shop and consenting technicians.
What the data can support, and what it cannot
Records can support task ranking by frequency, duration and parts, vocabulary grounding, diagnosis evaluation and retrieval. Video can support visual understanding, step recognition, progress estimation and pretraining.
They cannot support control. Human service video contains no robot joint states, commanded actions or torques; lug-nut removal with an impact wrench is invisible in force terms. Working with a vehicle's weight, wheel balance and fastener torque needs robot-side data, or instrumented tools plus later retargeting. Records also hide variation: a flat-rate time reflects an average, not what a given job required.
The data does not carry vehicle-specific procedure text either. For that, the OEM service information is the authority, and its access is a separate matter described on the OEM service documentation access page.
Supplier types
Dealer service departments hold OEM-aligned processes, repair orders by model and technician training. Rights questions include the OEM franchise agreement.
Independent repair chains hold higher volumes across makes, standardized procedures and sometimes in-bay cameras. Contracts differ by franchise.
Fleet maintenance operations (delivery, transit, rental) hold repeated service on a small set of vehicle types. The repair shop networks as suppliers page covers how to approach multi-location firms.
Tire, quick-lube and glass specialists offer very repetitive tasks and tightly scripted workflows.
Illustrative package
# Illustrative only - synthetic values, not an available dataset
package: vehicle_service_records_and_video_v0
records:
repair_orders: fields [ro_id, vehicle_year, vehicle_make_model_class, complaint, labor_op_code, flat_rate_hours, actual_hours, parts[], outcome, comeback_flag]
vehicle_identifier: VIN removed; vehicle class kept
parts_join: part numbers mapped to PIES-style attributes where supplier provides
video:
tasks: [oil_change, tire_rotation, brake_inspection]
views: technician head-mounted + 1 fixed bay camera
labels: step segments, tool in use, lift state, outcome
linking: ro_id <-> video_id
metadata: shop_009, technician_T14 (pseudonymous), bay_type, lift_type
split: held out by shop and by vehicle class
Standardization is both the draw and the trap
Standardized procedures are what make service data attractive, but they also hide variance that a robot meets in real vehicles: seized fasteners, aftermarket parts, accident repairs and corrosion in salt-belt regions. Records usually show the flat-rate time and the code, not the fight with a rusted bolt. Part drawings in an exploded view help here, because they show how a seized part sits in its assembly. Ask whether the supplier retains technician notes about difficult jobs, and whether actual time differs from book time in a way that can be analyzed.
Vehicle diversity matters more than clip count. A set built around three common models in one region will teach little about underbody layouts elsewhere. Request coverage tables by vehicle class, model year bands and region, and keep a held-out vehicle class for evaluation.
Quality and acceptance criteria
- Repair orders linked to video through an identifier with matching times.
- Vehicle identifiers and customer data removed or controlled according to the agreement.
- Parts mapped to a stated standard or a documented crosswalk.
- Coverage across vehicle classes the plan names, not one model repeated.
- Comeback or callback flags present if the buyer wants outcome labels.
- Video shows fastener and tool contact in the close view, not only the vehicle.
Permissions and provenance
Repair orders carry personal data: names, contact details, VINs, license plates and service history. Video can show customers, license plates and other technicians. Diligence questions include what notice technicians and customers received, what the shop's dealer or OEM agreements say about service data, and whether removing VINs and names is sufficient for the supplier's own policy. Redaction does not resolve permission issues by itself. Third-party content in procedures, such as OEM manual excerpts, needs separate rights review.
Questions to ask suppliers
- Which vehicle classes, models and service types are covered, and over what period?
- Are repair orders linked to video, and how?
- How are parts identified: OEM numbers, aftermarket numbers, or a standard such as PIES?
- Do records include labor-op codes, flat-rate and actual times, and comeback flags?
- What was removed from records: VINs, names, plates, locations?
- What notice and consent did technicians give for video and AI-training use?
- Do franchise or OEM agreements restrict use of service data?
- Can additional bays or shops be recorded under the same terms?
Discuss your requirements
If you need vehicle-service records or video, describe the service types, vehicle classes and permitted uses to SourceX. SourceX looks for US businesses that hold suitable material, checks the data and the supplier's licensing permissions, and follows up about potential sources. Any new recording is scoped case by case, and a request does not guarantee a match. The catalog entries for maintenance logs and field service maintenance records describe adjacent record types, and the labor time standards page explains flat-rate fields.
Describe the workplace demonstration data your robotics project needs
Sources
- Auto Care Association ACES and PIES standards — fitment and product-information data standards.
- MaintNet: A Collaborative Open-Source Library for Predictive Maintenance Language Resources — automotive logbook data and normalization tools.