Datacenter · Proven
Lights-Out Sentinel: Human-vs-Robot Discrimination
In robot-patrolled halls, it alarms only when the moving thing is a person.
Demonstrated on real public research data.
How it works
A walking human stirs the WiFi field with a messy, rhythmic pattern — swinging arms, stepping legs, and always breathing. A wheeled robot glides at constant speed with a clean motor hum and no breath at all. The system tells the two apart, so robots stop drowning the alarms.
In the physics: Human micro-Doppler is broadband and quasi-periodic (limb swing 1-3Hz with harmonics, gait cadence) and always carries a 0.1-0.5Hz breathing component; wheeled robots produce narrowband, constant-velocity Doppler with motor-vibration lines and zero respiration — separable signatures in the 0.1-40Hz band.
See it work
This is real pipeline output for Datacenter, replayed. The exact reading depends on your room.
Go deeper
How the field becomes an answer
How we prove it
Machine-vs-human discrimination already class A on our data; validate trajectory tagging on OPERAnet + own-hardware with a roving vacuum robot and human confederate, measure per-track label accuracy.
Who it is for
The only motion alarm that ignores your robots — 'a human stepped into Hall C at 03:14, and we knew it wasn't the tape bot' — sold to hyperscaler physical security operations.
runtime Runs today in LatentField 2.0 as presence
The matching solution
Complete first-room estimateCage & Plenum Sentinel6 nodes · ~$320 parts · 120 min Request kit, subscribe, or set up →Compatible DIY hardware options
120 min · advanced
240 min · installer