ITEA MOSIM: Motion Capture and 3D Avatar Modeling
Updated: Oct 23
Simulating a factory is straightforward. Simulating the person inside it is not.
ITEA MOSIM was an R&D initiative aimed at that gap. Mimic contributed motion capture for industrial simulation, modular motion authoring, and 3D avatar work that let digital workers perform automotive production tasks with believable movement rather than looping animation.
The output was not a film. It was data that another team's system had to consume, recombine, and run.
Table of Contents
Project Overview
Human Movement as Reusable Simulation Data
MOSIM set out to close the distance between how people actually move and how digital humans move inside simulation environments.
The automotive production floor is a demanding test case. A worker reaches into a chassis, shifts weight, switches arm position mid-task, then repeats the sequence with small variations all day. Simulation systems need that behaviour to be adaptable, not pre-baked.
Mimic captured those tasks and delivered them as modular units the MOSIM team could recombine, alongside the 3D avatars that carried them on screen.
The Challenge
Motion That Has to Be Recombined, Not Replayed
Standard animation libraries fail in simulation for a predictable set of reasons:
Clips are fixed length and fixed context
Transitions between actions look mechanical
Task variation requires a new capture every time
Locomotion and manipulation are authored separately and fight each other
Playback drifts from the engineering intent of the simulation
An industrial simulation needs a digital worker who can pick up a different part, from a different height, in a different order, without a new shoot. That is an architecture problem before it is an animation problem.
There was a second constraint. Whatever Mimic delivered had to run inside the MOSIM team's own system, through their Thrift pipeline into Unity, not inside a studio's preferred toolset.
What a Modular Motion Unit Is
A Modular Motion Unit, or MMU, is a self-contained block of motion behaviour that a simulation can call, sequence, and blend with other units at runtime.
Instead of one long recording of an assembly task, the task is broken into components the system can reassemble:
Locomotion, walking and repositioning
Reach and grasp
Lift and carry
Fine motor assembly
Arm position switching
Idle and posture holds
The value is combinatorial. A modest library of well-authored units covers a wide range of task sequences, which is what makes simulation scenarios cheap to iterate.
This is the same principle behind embodied agents generally, where behaviour is composed from controlled modules rather than generated end to end. Teams working on that class of system will find the AI avatar for robotics page a useful reference point for how Mimic frames embodied behaviour outside pure entertainment contexts.
The Solution
Capture, Author, Integrate
Capture planning and recording. Mimic facilitated high-fidelity motion capture sessions designed around the actual task list, covering locomotion and task actions inside a simulated car production environment.
Performer direction. The team worked closely with performers so every industrial movement was authentic, from heavy lifting through to fine motor assembly. Industrial movement has its own economy. A performer who has never done the job will overact it, and the data shows it.
MMU development. Captured motion was authored into Modular Motion Units and validated for clean playback.
Runtime integration. Units were delivered into Unity through the MOSIM Thrift pipeline, so the MOSIM team could run them inside their own system without adaptation work on their side.
3D avatar modeling. Avatar geometry was refined for likeness and proportion, meeting the visual standard the simulation required.
Texturing. Clothing and shoes were textured with material tuning for visual consistency, including branding elements such as logos and garment details.
Why the Visual Layer Mattered
Simulation Is Also a Communication Tool
A simulation is reviewed by people. Engineers, managers, and stakeholders watch the playback and form judgements from it.
When the digital worker wears generic grey clothing, viewers read the output as abstract. When the avatar wears the right garments, the right shoes, and the right branding, the same motion data reads as a plausible account of the production floor.
That is why texturing and avatar refinement were treated as part of the deliverable rather than polish at the end. Visual consistency is what lets a technical result travel to a non-technical audience.
Teams that later need to put that same digital human in front of customers rather than reviewers usually shift to a persona-driven build. That handoff is what Mimic AI Studio is structured for, where identity, voice, and behaviour are tuned as deliberate production choices.
Impact
A Movement Library the Simulation Could Actually Use
What MOSIM demonstrates about motion capture for industrial simulation:
Task motion captured at the fidelity the engineering required
Motion delivered as recombinable units, not fixed clips
Runtime integration into the client's existing Unity and Thrift setup
Avatars refined for likeness, proportion, and material accuracy
A visual layer that made technical output legible to stakeholders
The project ran through the motion capture and character pipeline at Mimic Productions, the Berlin studio that has handled Mimic's capture, scanning, and animation work since 2012.
Applications Across Industries
Modular motion data supports work well beyond automotive R&D:
Automotive assembly and production planning
Warehousing and logistics simulation
Robotics and human-robot collaboration studies
Ergonomics and workplace safety analysis
Industrial training and onboarding
Construction sequencing and site planning
Aerospace and heavy manufacturing
Digital twin environments
Automotive teams extending this thinking into in-cabin and vehicle-facing experiences will find the AI avatar for mobility page the closer fit. For a broader view of where digital humans land first across sectors, the Industries overview is the practical starting point.
Credits
Client: ITEA4
Project: MOSIM
Release: 2020
Creative Partner: N/A
Format: R&D / Industrial Simulation
Mimic's Contribution: Motion Capture, MMU Development, 3D Avatar Modeling, Texturing, R&D
Movement Is the Hard Part
Digital environments are easy to build convincingly. The person inside them is where simulation usually falls apart, because human movement carries weight, intent, and correction that generic animation does not.
Motion capture for industrial simulation solves that by treating movement as engineering data with a visual layer on top, captured from people who move the way the job actually moves.
Explore Mimic Minds AI Avatar Platform →
For further information and in case of queries please contact Press department Mimic Minds: info@mimicminds.com
From Motion Data to a Production-Ready 3D Avatar
A reusable workflow begins with the target environment and performance requirements. Real-time applications, simulation, film and research may need different model complexity, rig controls and output formats.
Modeling translates visual references into geometry, topology, materials and textures. The character must balance appearance with deformation quality and the performance budget of the intended platform.
Rigging connects the model to motion. Body hierarchy, facial controls, retargeting standards and naming conventions determine whether captured movement transfers reliably across tools.
Motion capture requires calibrated performers, sensors and coordinate systems. Teams should record clean source data and document sessions so corrections remain traceable.
Retargeting is not a one-click step. Differences in proportion, contact, facial range and clothing may require cleanup or authored animation to preserve the intended performance.
Validation should include extreme poses, locomotion, hand interaction, facial speech and platform playback. A model that looks correct when still may fail under real movement.
Interoperability matters for collaborative research and production. Document formats, scale, skeleton definitions, metadata and version compatibility so assets can move between systems.
Rights and data governance apply to source references, performer movement and final characters. Store approvals and control access to reusable capture assets.
A project page should report documented objectives and methods without inventing research outcomes. General technical guidance should be clearly separated from verified project findings.
Practical Takeaways
Design the asset for its target environment.
Treat topology, rigging and retargeting as one system.
Validate motion on the final platform.
Preserve provenance, formats and rights.
Related Mimic Minds Resources
FAQs
What was the ITEA MOSIM project?
MOSIM was an R&D initiative under ITEA, a European research and innovation programme, focused on simulating human movement inside industrial environments. Mimic contributed motion capture, modular motion authoring, and 3D avatar work, released in 2020.
What is a Modular Motion Unit?
A Modular Motion Unit, or MMU, is a self-contained block of motion behaviour that a simulation can call, sequence, and blend at runtime, rather than replaying a fixed animation clip.
Why use motion capture instead of hand-keyed animation for simulation?
Hand-keyed animation encodes an animator's assumption about how a task is performed. Capture records how it is actually performed, including weight shift, hesitation, and posture, which is what simulation analysis depends on.
How was the motion delivered into the client's system?
Motion units were integrated into Unity through the MOSIM Thrift pipeline, so the MOSIM team could run playback inside their existing environment.
Why did the avatars need detailed clothing and branding?
Simulation output is reviewed by people. Accurate garments, shoes, and branding make the playback read as a plausible account of the real production floor rather than an abstract test.
Can this approach apply outside automotive?
Yes. Any sector that simulates human tasks benefits, including logistics, robotics, ergonomics, training, and construction.
Does this replace on-site time and motion study?
No. It gives those studies a digital environment to run in, where variations can be tested without stopping a production line.
How large does an MMU library need to be?
It depends on the task range being simulated. The advantage of modular authoring is combinatorial, so a well-scoped library covers more scenarios than its unit count suggests.
Conclusion
ITEA MOSIM motion capture and 3D avatar modeling creates value when it is connected to a clear audience need, reliable information and accountable human oversight.
Comments