Active research Three programmes — internal builds
Experiments & tests
Labs

Where we keep the systems that aren't finished yet. Everything here is a live test — physics rigs, capture pipelines and traffic models built to answer one specific question. Some of it will ship. Some exists purely to prove what doesn't work.

Programmes
Three
Focus
Simulation
Capture
Markerless
Region
Finland
01

Getting a simulated car to behave like that car.

In testing Simulation

Generic vehicle physics gets you a car. It doesn't get you a specific car — the way one particular chassis loads up mid-corner, how its dampers settle over a broken surface, where the rear steps out and how far. We're building a pipeline that takes real vehicle data and tunes a simulated counterpart until the two agree, then holds that agreement across terrain the reference car was never measured on.

Video coming soon
Terrain sweepReference vs. sim overlayTelemetry capture
Programme
Stage
Correlation testing
Surfaces
Tarmac · Gravel · Snow · Mud
Target
Per-vehicle fidelity
Loop
Measure → tune → validate
What we're solving
  • A tyre model that stays honest when the surface changes underneath it, not just on dry tarmac.
  • Suspension and weight transfer matched against reference data rather than eyeballed.
  • Repeatable test routes, so two builds can be compared like for like.
  • Automated scoring — how far is this run from the real car, numerically.
Stack
Physics solverTelemetryTerrain authoringRegression harness
Maturity
ConceptTestingProduction
02

Motion capture without the suit.

In development Capture

Marker-based capture is accurate and expensive, and it changes how people move the moment you put them in the suit. We're developing a markerless system that reconstructs skeletal motion from ordinary synchronised cameras — no markers, no volume rental, no setup ritual. The work right now is accuracy: closing the gap to marker-based ground truth, especially on fast limbs, occlusion and contact.

Video coming soon
Multi-view rigSkeleton solveGround-truth comparison
Programme
Stage
Under development
Input
Synchronised multi-camera
Output
Retargetable skeleton
Benchmark
vs. marker-based
What we're solving
  • Calibration that survives a room being knocked about between takes.
  • Occlusion recovery — limbs that disappear behind the body or props.
  • Temporal stability, so the solve doesn't jitter frame to frame.
  • Clean export into standard animation pipelines without hand cleanup.
Stack
Multi-view geometryPose estimationTemporal solveRetargeting
Maturity
ConceptTestingProduction
03

The Finnish road system, made legible.

Prototype Application

Finnish traffic rules are precise, and precision is exactly what's hard to hold in your head at a junction in the dark. This prototype turns the rulebook — right of way, priority roads, signage, seasonal conditions — into something you can interrogate: situations rather than paragraphs. Built as a testbed for how far an app can go in explaining a road system instead of just quizzing you on it.

Video coming soon
Interface walkthroughScenario modeSignage reference
Programme
Stage
Working prototype
Scope
Finnish road rules
Model
Situation-first
Platform
Mobile
What we're solving
  • Right of way explained as a decision at a junction, not a clause in a document.
  • Seasonal reality — winter tyres, ice, darkness, wildlife — treated as first-class context.
  • Signage lookup that works from what you saw, not from what it's called.
  • Content structured so it can be kept current as regulation changes.
Stack
Rules engineScenario modelMobile clientContent pipeline
Maturity
ConceptTestingProduction
Collaboration

Interested in one of these, or have a harder version of the same problem?

Several of these programmes are open to partners, pilot data, or test subjects. If something here overlaps with what you're working on, we'd like to hear about it.

Get in touch