Introduction
Automotive plants provide some of the clearest public humanoid evidence because the site, task and production context can be named. The evidence still spans very different levels: Figure has published BMW operating hours and part counts, while other programs are described through pilots, commercial agreements or planned deployments. This page keeps those categories separate and records robot generation, plant, task, operating evidence and missing intervention or uptime data.
Key findings
- Figure reports Figure 02 loaded more than 90,000 sheet-metal parts at BMW Plant Spartanburg across more than 1,250 operating hours and contributed to production of more than 30,000 X3 vehicles.
- Figure 03 later returned to BMW for a different logistics sequencing workflow, so Figure 02 production metrics should not be assigned to Figure 03.
- Apptronik and Mercedes-Benz have a commercial relationship and factory testing record, but public fleet size, intervention rate and sustained production metrics remain limited.
- Automotive tasks should be compared using cycle time, first-pass success, interventions, recovery time and impact on the production line.
- Supplier announcements remain the main public evidence source, so company-reported metrics should stay attributed.
Humanoid Robots on Automotive Production Lines - evidence comparison
The table records what each source establishes and keeps missing data visible.
| System or method | What the evidence establishes | Evidence class | Main unresolved point |
|---|---|---|---|
| BMW and Figure | The strongest public evidence includes named plants, tasks and company-reported production metrics. | Documented production trial | Public evidence is concentrated in supplier announcements. |
| Mercedes-Benz and Apptronik | A commercial agreement and factory tests are public; exact fleet size and production metrics remain limited. | Commercial agreement and pilot | Comparable success rates, uptime and total cost are rarely published. |
| Schaeffler and Agility | Public agreement covers Digit deployment plans, with operational scope evolving over time. | Commercial agreement | Paid status and robot count are often undisclosed. |
| Chinese automakers and UBTECH | Multiple manufacturers announce Walker programs; robot counts and daily use require case-by-case verification. | Company-reported pilots and orders | Public evidence is concentrated in supplier announcements. |
Evidence classification
- Officially documented specifications, standards or project status stated by the responsible organization.
- Real-system evidence demonstrations or deployments performed on physical hardware under described conditions.
- Company claim a numerical or operational statement reported by the company and not independently audited.
- Simulation or research evidence useful for mechanisms, but not proof of field deployment.
- Insufficient public evidence control mode, trial count, version or operating conditions are missing.
Definition and deployment boundary
An automotive humanoid deployment is a documented activity inside a vehicle or component plant using a complete humanoid robot. It excludes conventional industrial arms, cobots, autonomous carts and demonstrations staged outside production.
How a factory workflow is engineered
Select tasks with stable parts, bounded reach and low consequence of delay. Calibrate robot perception to racks, bins and fixtures. Connect safe states to line controls without allowing an AI policy to bypass them. Measure takt time, first-pass success and intervention rate. Classify each customer relationship as test, pilot, paid pilot, deployment or announcement. Latency, calibration and safety limits can change the result even when the high-level model remains the same.
How to classify pilots and deployments
This page evaluates automotive factory robots humanoid at the level of the named system and dated source. Material from Figure AI, Apptronik, Agility Robotics is separated by task and test setting, with special attention to site, task, robot version, run hours, interventions, throughput and whether the activity was a test or production operation. Missing operating data is reported as missing, not converted into a maturity claim.
Operational failure modes
The main failure modes are concrete A task can meet cycle time only after fixtures are redesigned. Sheet metal and heavy parts increase pinch and drop hazards. Changing vehicle variants can invalidate a learned policy. Production claims may combine multiple robots and shifts without denominator data. A memorandum can be mistaken for a delivered fleet.
Tasks with credible industrial value
Credible applications include Part sequencing, rack transfer and machine tending, Inspection and data capture at human-accessible stations and Repetitive material movement between fixed points. These applications should be described with the robot, task boundary, operator role and environmental constraints. Experimental capability, commercial availability and routine deployment are reported as separate statuses.
Limitations and missing information
- Public evidence is concentrated in supplier announcements.
- Comparable success rates, uptime and total cost are rarely published.
- Paid status and robot count are often undisclosed.
- Specifications, prices, repositories and deployment status can change after publication.
- Benchmarks from different robots or environments are not directly comparable.
Conclusion
Automotive humanoid evidence is strongest when the plant, robot generation, task and operating denominator are published together. Figure’s BMW record provides measured factory exposure, while other programs still rely more heavily on pilot or agreement disclosures. Expansion decisions need intervention, failure and maintenance data that are rarely public today.
Frequently asked questions
How should automotive factory robots humanoid be evaluated?
Evaluate an automotive humanoid at the exact plant and task level. Record the robot generation, fixture and part geometry, cycle-time requirement, first-pass success, interventions, recovery, charging and line impact. A commercial agreement without those operating fields is not equivalent to production evidence.
What real-world evidence is available?
The strongest public operating record in this page is Figure 02 at BMW Plant Spartanburg, where Figure reports more than 90,000 sheet-metal parts loaded across more than 1,250 operating hours. Apptronik and Mercedes-Benz provide a separate commercial-and-pilot record with less public operating detail, so the two cases remain different evidence classes.
Sources and methodology
Sources for automotive humanoid deployments were rechecked on September 26, 2026 using current official material from Figure AI, Apptronik, Agility Robotics and UBTECH. Company-reported operating figures remain attributed, and tests, commercial agreements and production operation are labeled separately.
Related TechniaHQRobot guides
Source review notes
Review date recorded July 11, 2026
Supported by the cited sources
- BMW and Figure: The strongest public evidence includes named plants, tasks and company-reported production metrics.
- Mercedes-Benz and Apptronik: A commercial agreement and factory tests are public; exact fleet size and production metrics remain limited.
Not confirmed or incomplete
- Public evidence is concentrated in supplier announcements.
- Comparable success rates, uptime and total cost are rarely published.
- Paid status and robot count are often undisclosed.
Likely to change quickly
- Commercial availability, prices, model versions and software access.
- Deployment counts, company partnerships and repository maintenance status.
Share this article
Share the current TechniaHQRobot article page.
Continue reading
Open the latest robotics reporting, Physical AI analysis and hardware notes.