Disruptive Technologies

Default operation
The Wireless Desk Occupancy Sensor uses a combination of temperature measurements and
machine learning to determine if a desk is occupied or not based on changes in temperature
caused by the presence of people sitting at a desk. A desk occupancy event with an OCCUPIED
state is sent to the cloud when the desk becomes occupied. Similarly, a new event is sent to
the cloud with a NOT_OCCUPIED state when a desk becomes available.

The sensor will typically detect if a desk is occupied within 5-10 minutes of the person arriving
at the desk. Similarly, it will typically detect if a desk is not occupied within 5-10 minutes of the
person leaving.

The radio protocol used is SecureDataShot, and the data is relayed to DT cloud infrastructure
using a SecureDataShot enabled gateway, also known as a Cloud Connector. Data can be
viewed directly in Studio (web application) or sent to external services using webhooks or a
REST API

Heartbeat Interval
The Heartbeat Interval is a user configurable interval that controls how often the sensor reports
to the cloud that it is online and operational. The Wireless Desk Occupancy has a fixed 5 minute
heartbeat interval.

Responsiveness
Occupied: Up to 10 min (typical) Not Occupied: Up to 10 min (typical)

Accuracy
The datasets used to train the ?achine learning algorith?s have been collected fro?
sensors in a nor?al office building environ?ent (20-25°C, 15-60¡ RH). Given a 10
?inute delay, in si?ilar conditions, the following accuracy can be expected:

¬
? Probability of detecting OCCUPIED, when the desk is occupied: 98¡¬
? Probability of detecting OCCUPIED, when the desk is not occupied: 2¡¬
? Probability of detecting NOT OCCUPIED when the desk is not occupied: 99
? Probability of detecting NOT OCCUPIED when the desk is occupied: 1%


While the standard latency for detecting so?eone sitting down or leaving is
approxi?ately 10 ?inutes, variations in environ?ental conditions and specific
scenarios ?ay result in longer detection ti?es.


For ?ore infor?ation about the expected accuracy, contact Disruptive Technologies.

Important:

The ?achine learning ?odel used to deter?ine if a desk is occupied or not is trained based
on data fro? a typical office environ?ent (20-25°C, 15-60¡ RH). While the sensor can be
used in environ?ents outside this range, the detection accuracy ?ight be affected. DT
continuously i?proves the ?achine learning ?odel to cover a broader range of
environ?ents.