Short answer
Standard work documents and job breakdown sheets give a humanoid program the structure of a manufacturing task: ordered steps, timings, key points and the reasons behind them. They hold no motion trajectories, forces or camera views, so they work as labels and scripts for linked video, not as control data. Source them from manufacturers and industrial engineering groups, and verify file ownership.
The buyer's problem
A plant that has run lean programs for years may hold a library of standard work sheets, job breakdown sheets, combination tables and instruction cards. They look like a ready-made description of tasks a humanoid might one day perform. The buyer's problem is deciding what that paper can actually contribute, and what has to come from recordings, robots or instrumented capture.
This page covers the documents themselves. For the broader limits of written procedures, see what SOPs can and cannot teach a humanoid. For the manufacturing cluster as a whole, start at the manufacturing and assembly data hub.
What these documents contain
Names vary by plant, and the family is broadly similar. Check which of these a supplier actually holds:
- Standard work sheets: the sequence of operations at a station, the target cycle time and the work-in-process an operator should hold.
- Job breakdown sheets: a task split into steps, often with the key points that make a step succeed and the reasons for them. Sheets in the Training Within Industry job-instruction tradition are commonly organized around steps and key points (confirm the method's details against the source), so confirm whether a supplier's sheets carry the "reason" column or only step names.
- Standardized work combination tables: manual time, machine time and walking time laid out against a time scale for one operator.
- Motion-element analyses: tasks decomposed into basic motions with assigned times. The MTM Association, a non-profit founded in 1962 with more than 240 corporate members, describes Methods-Time Measurement as a method standard for describing, evaluating and designing human work, analyzing manual operations into basic motions such as reach, grasp, move, position and release (MTM Association).
- Operator instruction cards and change histories: the revision trail showing why a step changed.
Industrial engineering and EHS consultancies may also hold analyses they produced for clients, including posture ratings and timing studies. Those sit in a different rights position from a plant's own paperwork.
What the data can support, and what it cannot
| Use | Standard work documents can help because | What else you still need |
|---|---|---|
| Task decomposition and step ordering | Steps are already named and sequenced by people who run the line | Video or robot data to ground each step |
| Success criteria and key points | Key-point columns state what makes a step correct | Observations showing what correct looks like |
| Language for instruction-paired data | Step text can seed instruction sentences | Annotator review; video of each step |
| Timing priors and cycle-time baselines | Targets and element times exist | Measured times from your own station and embodiment |
| Evaluation scoring rubrics | Steps become checklist items | A held-out evaluation design |
The same table in negative form matters more. These documents do not contain robot actions, joint states, end-effector trajectories, contact forces, camera calibration or object appearance. A motion-element label such as "position" tells you a human did something at a point in a work cycle; it does not specify a path, a grasp or a compliance strategy a humanoid could execute. Treat them as structure and language, never as a physical-control demonstration. For why documents and demonstrations are different assets, see what human demonstration video can and cannot do for robot learning.
Linking the documents to video
The documents become far more useful when each step maps to a time range in a recording of the same station. That linkage is where most of the preparation work sits. Time and motion study recordings are the natural companion asset, because a study video may already be organized around the elements on a combination table. Similarly, assembly sequence data shows how the ordering information is structured when it comes from engineering systems rather than shop-floor paper.
Expect these joins to be manual. Ask whether step identifiers on the sheet match anything on the video file, such as a station code or revision number, or whether an annotator will need to align them by eye.
Illustrative package: standard work linked to station video
# Illustrative only - synthetic values, not an available dataset
package: standard_work_with_video_v0
station: station_A07 (synthetic)
documents:
standard_work_sheet: pdf, revisions r1-r4, effective dates retained
job_breakdown_sheet: structured table (step_id, important_step, key_point, reason)
combination_table: spreadsheet, manual/machine/walk time per element
links:
step_to_video: step_id -> (clip_id, start_s, end_s), reviewed by second annotator
revision_to_video: which sheet revision was in force on the recording date
metadata:
operator_id: operator_A12 (pseudonymous)
document_language: en
time_basis: stopwatch / MTM-coded / mixed (stated per element)
excluded: robot actions, joint states, forces, camera calibration
rights: documents licensed by plant; any consultancy-authored analysis flagged separately
Change the fields to fit your architecture. A team building instruction-paired data will care most about the reason text and revision linkage; a team building progress or reward estimators will care most about the step time ranges.
Limitations and additional preparation
- Idealized work. A standard describes the intended method. Floor practice drifts, and the sheet may lag behind. Comparing sheet to video is itself a quality check.
- Variation hidden. Standards rarely record recoveries, rework or mis-seated parts, which are the cases a robot meets.
- Timing is not motion. Element times describe human performance and say nothing about humanoid reach, payload or speed limits.
- Redaction. Part numbers, torque specs, supplier names and customer drawing references often appear on the sheets. Locate and redact or exclude them; see manufacturing data confidentiality.
- Format. Many plants keep sheets as scanned images or presentation slides, so extraction work may be needed before the text is usable.
- Provenance. Sheets often derive from customer or corporate templates; confirm who authored each one.
Permissions and provenance
Employee video linked to the documents raises its own questions; see worker consent and identifiable people in workplace recordings. Consultancy-authored time studies may belong to the consultancy or to the client under the engagement terms. Parent-company templates, customer-specified work instructions and published method standards can each carry restrictions. This is not legal advice; settle ownership in writing before pricing.
Questions to ask suppliers
- Which document types exist per station: standard work, job breakdown, combination table, instruction card?
- Do job breakdown sheets include key points and the reasons for them, or only step names?
- How many revisions exist, and can each revision be tied to a date range and to recordings?
- Is there video of the same station and shift, and are step identifiers shared with the video files?
- Were element times measured on the floor, taken from a predetermined-time system, or copied from a template?
- Who authored the analyses: the plant, a consultancy or a customer, and who can license them?
- Do the sheets contain customer part numbers, drawings or specifications that must be removed?
- What file formats and languages are used, and how much is scanned?
- Are deviations, rework loops and abnormal conditions documented anywhere?
Discuss your requirements
If your program needs task structure paired with recordings, describe the stations, document types 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. A request does not guarantee a match; SourceX follows up about potential sources. See the humanoid training data guide for the wider map, and the catalog entries for SOPs, playbooks and knowledge bases and manufacturing quality records for related document types.
Discuss your humanoid training-data requirements with SourceX
Sources
- MTM ASSOCIATION e. V. (Methods-Time Measurement) — MTM Association's description of Methods-Time Measurement and basic-motion analysis.