One core, three real-world scenarios
The same Physical World Perception Model, mapped onto different environments. Three illustrative scenarios show how the model reconstructs events from sensing — and then anticipates what comes next.
Care & Healthcare
For care facilities and people living alone, the model continuously senses room state and prompts staff when needed. Ambient, camera-free, built to run for the long term.
Illustrative scenario · demo data
Late at night, room 302's bathroom dwell exceeds its threshold — the dashboard raises an alert and recommends a check.
- 02:11Bathroom entry triggered, room 302
- 02:11Continuous presence in bathroom
- 02:35Dwell 24 min; night threshold 15 min
The model reconstructs repeated triggers into one continuous event: the resident entered the bathroom at 02:11 and has stayed beyond the night-time safety threshold.
Classified as an over-threshold bathroom dwell alert — a room check is recommended, and if unacknowledged for 5 minutes it escalates to the on-call phone.
Robotics & Physical AI
A robot knows what it can see, but not what's happening elsewhere. The model supplies environmental context beyond its field of view, plus spatial history.
Illustrative scenario · demo data
The model detects someone resting in the bedroom and advises the robot to hold off entering that zone.
- 14:12Kitchen activity signal ends
- 14:15Movement along the corridor
- 14:18Stationary presence in bedroom
The model reconstructs the spatial state: a person moved from the kitchen through the corridor into the bedroom, which is now occupied and at rest.
Anticipating the bedroom isn't suitable to enter right now, the model advises the robot to hold or reroute until the spatial state updates.
Security & Smart Buildings
Without identifying anyone, the model judges how space is really used and where abnormal states arise — across server rooms, warehouses, archives and unattended areas.
- Server Rm AAfter-hours activity
- Warehouse BDwell 38 min
- OfficesNormal use
- ArchiveUnoccupied
Illustrative scenario · demo data
After hours, abnormal activity in the server room — the model correlates signals across zones.
- 21:40Access signal, Server Rm A
- 21:43Ongoing activity inside
- 21:48After hours · no shift on record
The model reconstructs access and interior activity into one event: someone entered the server room after hours and remains active, with no matching shift.
Classified as after-hours abnormal activity — security is prompted to verify, while adjacent-zone signals are checked for any linked trajectory.