The initial approach was to only look at the air quality in the room, but it was found that reasonably priced air quality sensors have a tendency to saturate over time so it wouldn’t be reliable in the long run and the pricier versions would be too big of an investment for a large building with many rooms.

From this a new approach was adopted which is to build a model of how the air quality changes with the occupancy of the rooms. The number of people in the room can be counted with a pair of time of flight sensors in the doorframes then the air quality sensor only needs to be used periodically to recalibrate the model. This approach allows for continuous course correction of the system if the conditions of the room change over time.

The model would be applied to the data in the cloud and the analytics can be accessed by the users, building managers from anywhere and the output can be linked to HVAC systems to automatically adjust the settings to prevent the decline in air quality based on the expected change as well as maintain an efficient control on heating and cooling.

This solution would be unique as instead of being reactive it is preventive which means the air quality never declines below comfort levels whereas in current systems the quality has to be out of the comfort level for the system to adjust itself.

Below you can find how the number of people directly affect air quality. Especially noteworthy is the delay - as the number of people changes, so does the air quality but it takes over an hour for the effects to be visible!

We primarily used a pair of TOF sensors (monitoring people entering and leaving the room) and a simple air quality sensor. At the begining we experimented with a ultrasound-based sensors but they were not precise enough to provide sufficiently accurate results for people counting.