Short answer
Wire harness assembly is commonly manual work between machine wire processing and final test: populating connector housings, routing wires on a form board, bundling with tape, ties or sleeves, labeling and checking. Harness manufacturers hold drawings, work instructions, test results and some station video, and IPC/WHMA-A-620 defines acceptability. Video does not show crimp quality, latch feel or how wires deform under tension, so robot or instrumented data is needed.
The deployment problem
Harness work is a frequent candidate for humanoid-style automation because it tends to stay in human hands. Every harness is a different product, the objects are long and flexible, and the operations that matter happen in small spaces: a terminal entering a cavity, a wire turning at a peg, a tape wrap closing a branch. Fixed automation struggles with that variety, which is the argument for a general-purpose robot.
For a data buyer, harness plants offer something unusual: each build is documented to the wire. The question is what that documentation, plus whatever video exists, can teach, and what a robot still has to learn from its own data.
Task anatomy
Wire cutting, stripping and crimping are often done on machines, so the manual task usually starts with kitted, terminated wires.
| Step | Objects | What the operator does | Typical failure modes |
|---|---|---|---|
| Kit | Cut wires, connector housings, seals, clips, tape, sleeves | Match the kit to the build sheet | Wrong wire; missing seal |
| Populate connectors | Terminated wire, housing cavities | Insert each terminal into the cavity named in the wire list until it engages, then confirm with a tug | Wrong cavity; terminal not latched; bent terminal |
| Lay out on the form board | Board with pegs and clips, drawing | Route each wire along the drawn path, keeping lengths and branch points | Wrong path; crossed branch; slack |
| Bundle | Tape, ties, sleeves, clamps | Wrap or fasten at the positions the drawing shows | Gaps in tape; tie too tight; sleeve short |
| Add accessories | Clips, grommets, labels, covers | Place at specified positions and orientations | Misplaced clip; wrong label |
| Test | Continuity or electrical test fixture | Mate the harness to the fixture and read the result | Open or short; wrong pin mapping |
| Inspect and pack | Drawing, acceptance criteria | Accept, rework or reject | Missed defect |
What the acceptance standard contributes
IPC/WHMA-A-620 covers materials, methods, tests and acceptability criteria for crimped, mechanically secured and soldered interconnections and related assembly of cable, wire and harness assemblies. IPC's store lists Revision F as the current revision, notes that Revision E was released in 2022, and says the standard defines three product classes, with the user responsible for choosing the class (IPC, 2025).
Three consequences for a buyer. First, the acceptance criteria are effectively label definitions for pass and fail, so record the revision and class applied to every build. Second, the standard is a purchased document; a supplier can refer to it but the dataset should not be expected to carry its text. Third, the standard is reported not to set inspection frequency or cross-section and X-ray criteria, so those come from the plant's or customer's own rules. Ask which apply if you need defect labels that rely on internal inspection.
Which data feeds which part
Layout and wiring intent. Harness drawings, form-board layouts, cut lists and wire lists with cavity assignments are structured documents that give the intended route and connection map. They are a strong source of task structure and success criteria, and they are usually customer-owned engineering data; see schematics and wiring diagrams and the catalog entry for CAD and PCB engineering files.
Hand work. Overhead video of the form board shows routing order and bundling. A head-mounted or wrist camera shows terminal insertion and regrips. Station cameras that look at the board from above rarely resolve a cavity, so insertion needs a close view.
Outcome labels. Electrical test logs give a pass or fail per serial, and photos taken at inspection give visual labels. Their use as training signals is covered in cable harness and connector inspection.
Pace and sequence. Standard work sheets and time studies give element order and durations; see assembly sequence data.
What this data can and cannot support
| Asset | Supports | Cannot support alone |
|---|---|---|
| Drawings, wire lists, cut lists | Task structure, route targets, cavity assignment, success criteria | Motion and perception |
| Overhead form-board video | Routing sequence, bundling order, step timing | Insertion feel, tension, tape application |
| Head-mounted or wrist capture | Terminal insertion, regrips, wire handling | Forces; robot actions |
| Electrical test logs | Outcome labels per harness | Which hand action caused a fault |
| Inspection photos | Visual defect labels | Crimp internals, latch engagement |
Wires are deformable linear objects: their shape depends on gauge, insulation, length and how they were bent before. Video of a person laying a wire shows the path but not the tension or the corrections that kept it there. Teaching a robot to do the same needs robot or teleoperation episodes, as discussed in deformable object handling. Seating a terminal is a contact problem. Contact-rich robot datasets such as RH20T pair every sequence with visual, force, audio and action data (Fang et al., 2023), which is the kind of signal harness archives lack. See connector mating and precision insertion and force/torque data for what to specify.
Capture and labels
For commissioned recording, use an overhead camera for the whole board, a head-mounted camera for hands and a close fixed or wrist view at the point of insertion. Mark the board with an ID visible in frame. Ground truth comes from the wire list: each insertion event can be checked against the expected cavity. A request to SourceX should say which harness families, connector types and views you need.
# Illustrative only — synthetic harness, wire IDs and values
build_id: HX_site044_2210
harness: family_door_module_B (customer-owned design: approval attached)
acceptance: {standard: IPC/WHMA-A-620, revision: F, class: recorded per build}
cameras: [board_overhead, head_mounted, insertion_closeup]
events:
- {step: populate, wire: W014, connector: J2, cavity_expected: 6, cavity_observed: 6, tug_test: pass}
- {step: populate, wire: W015, connector: J2, cavity_expected: 7, cavity_observed: 9, tug_test: n/a, error: wrong_cavity, corrected: true}
- {step: route, branch: BR2, pegs: [P3, P4, P7], slack_mm_flag: false}
- {step: bundle, method: tape_spiral, region: BR2, gaps: 0}
test: {continuity: pass, fixture: TF_12}
operator_id: operator_C09 # pseudonymous
excluded: [crimp force, insertion force, wire tension, robot actions]
Coverage should span connector families, wire gauges, branch counts, harness lengths and operators. A set built from one harness design shows one route and one set of connectors.
Supplier types
- Harness manufacturers. Hold drawings, work instructions, test logs and sometimes station video; most build to customers' designs.
- OEM plants with in-house harness lines. May own both the design and the footage; see contract manufacturers as data suppliers for the difference.
- Vocational and training programs. May hold recorded harness courses, with learner-consent questions; see vocational training and certification recordings.
- Test-equipment vendors. Hold fixtures and test logs, usually on behalf of their clients.
Limits for this task
- Every harness differs. Routes and connectors are product-specific, so generalization needs many designs.
- Hidden quality. Crimp quality is inside the terminal and is checked by pull tests or cross-sections, not by video.
- Hidden mating. Latch engagement is felt or heard, rarely seen.
- Occlusion. Fingers cover the cavity at the moment of insertion.
- Pace. Skilled operators work faster than a robot is likely to.
- Design ownership. Drawings and part numbers belong to customers.
Questions to ask suppliers
- Which harness families, connector types and wire gauges would recordings cover?
- Do any cameras see cavities during insertion, or only the board from above?
- Are wire lists with cavity assignments available, and who owns them?
- Which IPC/WHMA-A-620 revision and product class applied to each build?
- Are test results linkable to serial numbers and to video?
- Are rework and wrong-cavity corrections kept?
- Do any customer designs carry export-control or defense-related restrictions?
- What notice or consent covered operators, and does it include licensing to an outside company?
Permissions and provenance
Harness drawings, part numbers and labels belong to the customers who designed the product, and some designs carry export-control or defense-related restrictions that limit who may see them. Operators appear in close-up. See manufacturing data confidentiality and worker consent and identifiable people. Masking part numbers does not by itself settle permission. This is not legal advice.
Discuss your requirements
If you are scoping a harness program, tell SourceX the harness families, views, labels and permitted uses you need. SourceX looks for US businesses that hold suitable material, checks the data and the supplier's licensing permissions, and coordinates the license and delivery. If no existing asset fits, SourceX can scope a commissioned collection of new recordings. A request does not guarantee a match; SourceX follows up about potential sources.
Start from the manufacturing and assembly hub or the humanoid training data guide. The catalog entry for manufacturing quality and inspection records and the manufacturing industry page give wider context.
Describe the workplace demonstration data your robotics project needs
Sources
- IPC/WHMA-A-620F Requirements and Acceptance for Cable and Wire Harness Assemblies — acceptability criteria and product classes for harness assemblies.
- RH20T: A Comprehensive Robotic Dataset for Learning Diverse Skills in One-Shot — force, audio and action data in contact-rich robot sequences.