How it works
Cameras on the trucks record the site as they drive. The platform builds a 3D map from it, keeps it current, and reads inventory and traffic off it.
Perceive
Cameras on working trucks pick out what's in the aisle: pallets, boxes, columns, doors.
Locate
Every frame is placed in the building's own coordinates, so what was seen has an address.
Remember
Each pass adds to the same model, so positions and movements persist as history.
Understand
The map answers in the operation's own terms: which bay, which pallet, how long it has stood there.
The map
Three layers, each answering a different question and stored differently. The semantic layer gets its coordinates from the localization layer. The dynamic layer captures how semantic content changes over time. On top sits the digital twin, the explorer where they come together and the simulator that runs the site forward.
Localization
Where each forklift is. A metric 3D reference of the site that each forklift localizes against using its camera.
- Live position
- Routes driven
- Site map
Semantic
What is there. The racking, every storage position, and every pallet, box and other object seen during a drive, placed in 3D on the site map.
- Racking layout
- Storage positions
- Objects
Dynamic
What changed, and when. Per storage position, whether it was full or empty and when it was last seen, what changed between passes, and the pickups and dropoffs that add up to trips and flows.
- Fill per position
- Shelf changes
- Trips and flows
Every drive updates all three layers, at different paces. Localization gives the semantic layer a position for every camera frame. The semantic layer gives the dynamic layer a stable set of storage positions and objects to watch over time.
Layer
Update step
Artifact
Delivered to
Dynamic
Observe and compare
timestamped per storage position, compared with the last pass
Fill, changes, trips
nothing overwritten
WMS
continuous
Semantic
Fuse into the scene
depth and detections into voxels, layout, and 3D objects
Updated scene
same storage ids
Spatial Explorer
minutes after a drive
Localization
Align, merge, measure
aligned to the reference and merged, scored across drives
Approved map version
reference stays fixed
Forklifts
only after it scores better
The technology
The device on the forklift handles what has to happen now. Heavier models and the full history run in the cloud, or on your own servers when a site requires it. Every step is a module, so where it runs is a setting, not a rebuild.
On the device
Position is computed on board, so it holds through WiFi gaps. Only positions and events leave the forklift in real time. Video is never streamed, and recordings upload whenever the forklift has connectivity.
- Cameras and motion sensor
- Privacy masking
- Positioning
- Light-weight detection
- Recording
In the cloud
The uploaded recordings feed the map and the analysis, next to the heavier models and the full history. Map updates flow back to the forklifts.
- Map building
- Semantic understanding
- Precision detection
- Analysis
- Portal and WMS export
On premises
The same components on servers inside the building, for sites where recordings may not leave the network. The portal and the exports are unchanged.
- Cloud components on your servers
- Recordings stay in your network
- Same portal and exports
- Per site
Configurations
Since every module can be switched per vehicle, the hardware question comes down to three configurations.
Configuration A
Camera box
The smallest and cheapest box. Position and inventory are available after upload rather than live. A fit for manual forklifts where live tracking is not required.
On the device
CamerasPrivacy maskingRecordingPositioningDetectionSlot occupancyPayload eventsConfiguration B
Localizing box
Live position on the map, unaffected by WiFi gaps. All inventory and flow analysis happens in the cloud from the recordings.
On the device
CamerasPrivacy maskingRecordingPositioningDetectionSlot occupancyPayload eventsConfiguration C
Perception box
Slot occupancy and payload events computed on the vehicle in real time, with only small updates leaving it. Requires more onboard compute, which automated vehicles typically already have.
On the device
CamerasPrivacy maskingRecordingPositioningDetectionSlot occupancyPayload events
Compliance
ISO 27001 certified, GDPR by default, hosted in the EU
ISO 27001
Certified to ISO 27001. Security controls are documented, monitored, and independently audited.
GDPR
Compliant with GDPR. Personal data is minimised by design and processed only for documented purposes.
EU hosted
Hosted within the EU. Customer data is stored and processed in the region, under EU jurisdiction.