The recent pandemic crisis compelled companies to monitor the number of people on their premises and make sure there aren’t too many people at the same time. IoT solutions that implement automatic people traffic and flow measures are the easiest, quickest and safest way to solve this challenge without additional manpower. They measure the number of people - already at the entrance.
But about going one step further and using those same tools to generate actionable insights on the usage of a building, allowing for better planning of the rooms and floorspace throughout the whole year?
This is the approach that the Zurich University of Applied Sciences (ZHAW) has adopted.
The ZHAW is one of the leading universities for applied sciences in Switzerland. In research and development, the ZHAW Zurich University of Applied Sciences orients itself towards central societal challenges - with a focus on energy and social integration. With its locations in Winterthur, Zurich and Wädenswil, the ZHAW is regionally anchored and cooperates with international partners.
With akenza, ZHAW conducted a POC in which the number of people in classrooms was measured continuously and analyzed to better understand the real usage of the rooms. This would allow improving the space planning and room utilization of the building. A fundamental requirement for this project was the correlation between the actual occupancy data reported by the sensors and the reservations made via the ‘Evento’ room booking system of the school. The purpose of this comparison, requested by the ZHAW, is to understand better whether a room is occupied by formal (a lecture for instance) or informal events (spontaneous student gatherings over lunch).
Goals of the people counting solution
The project aimed to analyze the real occupancy of the rooms in the buildings. This includes:
- Knowing how many people enter and exit the selected rooms.
- Obtaining the real occupancy per hour in correlation with the booking system ‘Evento’ of the school.
- Identifying the purpose of the room occupancy, i.e. official vs. unofficial.
- Check the efficiency and accuracy of the booking system.
Ideally, this project would be monitoring the spaces for an extended period, optimizing the usage of the rooms over time, possibly by repurposing the rooms depending on the month or season in correlation with the university's calendar
Architecture of the proof of concept
To find the ideal solution, consuming minimal resources, the project started with 4 people counting sensors, disseminated in three rooms (one room has two entrances). See the architecture diagram below:
- Devices: Xovis PC2S People Counting Sensor - GDPR compliant
- Connectivity: HTTPS - Encrypted HTTPS connectivity from akenza
- IoT Platform: akenza - IoT platform for connection management, device configuration and data aggregation / processing
- PowerBI - Microsoft Business Intelligence application for data processing and visualization
- Evento - Room booking software for booking a room at the ZHAW
The Xovis sensors count the number of people entering and exiting the classrooms, making up one of two data sets. The sensor data collected in akenza is connected to Power BI through REST API. Besides, the booking system for the rooms feeds the additional relevant data into the algorithm. In this POC, it is simply imported from an excel report stored in a SharePoint folder. The data is also forwarded to Azure in order to build a data lake, allowing for further analysis and visualization options.
This project was first tested in June 2021, with the pilot stage lasting about six months. Together with akenza and Valorando who developed the Business Intelligence Solution on Power BI, ZHAW was able to set up, test, and adapt the POC to generate the first learnings.
Why is akenza a good fit for this people counting solution?
Xovis sensors work seamlessly with akenza by integrating these via the Device Type Library (see below). In addition, akenza offers the flexibility to scale the solution efficiently. This allows ZHAW to start small, focusing on the data analysis part and scale later to potentially hundreds of devices.
Within the logic block, the series of received messages reporting every "event" (one person as in / out) are aggregated to form the effective occupancy count (number of people in the room) and passed on to the data model. In the case of the room with two entrances, data coming from two sensors are unified into one “virtual sensor” to obtain the effective occupancy even if a person enters from one door and exits from the other one. The data is then processed and the real-time occupancy of the room is calculated. Note that the counter resets every night. This is due to the 'ghosts,' people who haven't been recorded entering or leaving, throwing off the calculation if accumulated (a well set-up Xovis sensor grants a ghost-ratio under 1%).
Privacy concerns that are often raised in regard to people counting
The Xovis sensors contain a little camera, which concerns many in terms of privacy. It is essential to acknowledge that these cameras do not record an image but rather a ‘silhouette’ of data of some sort. Furthermore, the data is processed locally (within the device). Hence, there is never an actual image/video recorded and sent outwards. The entire process does not capture personal data and is anonymous, completely compliant with the EU General Data Protection Regulation.
The POC allowed validating the current system architecture and generating the first insights on the usage of the rooms at the ZHAW.
Over the following months, the project will be scaled to further ZHAW sites and adapted to a large device fleet. The akenza platform is able to scale to thousands of sensors based on the same IoT infrastructure. Regarding the data analysis layer, the data processing will be performed in Microsoft Azure, keeping Power BI for reporting purposes only.