Technology

What it is built from,
and what is proven.

Two questions an engineering team asks first. Both answered from the platform's own registry and its validation centre — not written for this page.

The evidence table below is deliberately unflattering. Every capability carries what would have to happen next, and for most of them that is a physical test that has not been performed. A page showing only what works would not be worth reading.
11
Modules
16
Capabilities
8
Implemented
4
Simulated
3
Partial
1
Not started

Capability evidence

16 capabilities, each with its implementation status and the next step that would validate it. Nothing here has run on physical hardware, which is why so many rows say so.

IDCapabilityStatus Next validation stepImplemented in
CAP-001 Vehicle Perception
Recover geometry of a vehicle surface from sensor input, well enough for downstream planning to act on.
Implemented Physical test against a real vehicle with a real sensor. Nothing in this repository has seen either. perception/src/scene.py, perception/src/pipeline.py
CAP-002 Vehicle Scanning
Drive a sensor over a surface and collect a usable sample set.
Simulated Bench test with a physical depth sensor. perception/src/sensors.py, perception/src/calibration.py
CAP-003 Panel Recognition
Separate a panel from its surroundings in the recovered geometry.
Implemented Physical test. Segmentation is currently geometry-only and has never met paint, reflections or ambient light. perception/src/segmentation.py
CAP-004 Digital Twin Generation
Maintain a model of the cell that mirrors measured state.
Partial HIL run with a physical robot reporting real joint state. robotics/src/state.py, robotics/src/cell.py
CAP-005 Surface Reconstruction
Fit a continuous surface to the sampled points.
Implemented Bench test against a measured reference artefact. perception/src/reconstruct.py, perception/src/quality.py
CAP-006 Wrap Planning
Turn a surface model into an ordered set of application passes.
Implemented Physical test - film behaviour is not modelled. robotics/src/process.py, software/patent_01/src/selection.py
CAP-007 Robot Trajectory Generation
Produce a time-parameterised, reachable, collision-free path.
Implemented HIL execution on a physical controller. robotics/src/trajectory.py, robotics/src/kinematics.py
CAP-008 Robot Execution
Drive a robot through a planned trajectory.
Simulated HIL, then bench. No physical robot has ever been driven by this software. robotics/src/bridge.py, robotics/src/cell.py
CAP-009 Material Handling
Account for film supply, liner and consumption.
Partial Bench test with real film on a real roll. robotics/src/process.py
CAP-010 Inspection
Sweep completed work and report quality.
Implemented Physical test against known-good and known-bad panels. software/patent_01/src/inspection.py
CAP-011 Defect Detection
Identify a defect in completed work.
Simulated Physical test. Detection currently runs against simulated defects that the same codebase injected, which cannot demonstrate real-world sensitivity. software/patent_01/src/inspection.py, software/patent_01/src/vision.py
CAP-012 Automated Correction
Return to a flagged defect and rework it.
Simulated Physical test. software/patent_01/src/control.py
CAP-013 Humanoid Panel Handling
Remove, carry and refit a panel using a humanoid platform.
Not started Implement it. There is no humanoid kinematic model, no hand model and no panel-handling routine in this repository; the rendered demonstrations are cinematic work, not evidence of platform capability. —
CAP-014 Workcell Coordination
Sequence the stages of a job across the devices in a cell.
Implemented HIL with more than one physical device. robotics/src/cell.py, integration/orchestrator.py
CAP-015 End-to-End Vehicle Workflow
Intake to release for a whole vehicle.
Partial Pilot. The end-to-end run exists only for a single panel; a whole vehicle has never been run end to end, in simulation or otherwise. integration/orchestrator.py, pilot/src/runner.py
CAP-016 Enterprise Integration
Exchange jobs and status with factory systems.
Implemented Integration test against a real MES or ERP instance. Adapters have only ever spoken to the in-repository fake. enterprise/src/adapters.py, enterprise/src/api.py

Module architecture

11 modules, with what each consumes, produces and depends on. Dependencies point inward: the robotics layer does not know the enterprise layer exists.

ModuleWhat it doesInputs OutputsDepends on
Surface treatment implementation
software/patent_01
Vision, magazine, selection, control, inspection, fleet and environment code organised against the claim elements of one granted patent. SurfaceModel, Environment WrapPlan, InspectionResult integration
Vehicle lighting implementation
software/patent_02
Code organised against the claim elements of one granted patent. — — integration
Service centre orchestration implementation
software/patent_03
Intake, dispatch, twin, custody, charging, geofence, utilisation, customer, security and explainability code organised against claim elements. — DigitalTwin integration
Dermatological imaging implementation
software/patent_04
Code organised against the claim elements of one granted patent. — — integration
Canonical data models
integration
Shared schema every other module speaks. — SurfaceModel, WrapPlan, RobotTrajectory, InspectionResult —
Robotics execution layer
robotics
Kinematics, trajectory generation, collision detection, singularity handling, tool changing, safety supervision and telemetry. WrapPlan, PanelGeometry RobotTrajectory, Telemetry integration, software/patent_03
Enterprise integration layer
enterprise
Jobs, inventory, adapters, event bus, audit, scheduler, digital thread and the versioned API gateway. Job, WrapPlan QualityRecord, DigitalThread robotics, integration, software/patent_03
Benchmark and ROI engine
benchmark
Manual baseline comparison, economics, ROI, sensitivity. Telemetry, BenchmarkConfig BenchmarkReport, ROIReport enterprise, robotics
Product licensing and activation
product_licensing
Licences, entitlements, Ed25519-signed activation, offline cache, updates, calibration records, acceptance testing and tenancy. License ActivationToken, UsageReport enterprise
Perception layer
perception
Simulated sensors ray-cast a scene; segmentation, surface reconstruction, scan quality and registration then run on the resulting point cloud WITHOUT access to the ground truth. Sensing is simulated; the reconstruction is real work on that data. SensorFrame, PointCloud ReconstructedSurface, ScanQuality, RegistrationResult robotics, integration
Cinematic visualisation
../autowrap_cinematic
Blender scene construction, wrap material system, shot library. Lives in a separate repository. RobotTrajectory RenderedFrames —
Every module is independently licensable. An unlicensed module is not loaded at all rather than hidden behind a flag, and every grant and refusal is audited.