DIGITAL SUSTAINABLE TRANSFORMATION
Ubiqisense has created a sensor that reveals how many people are in a building, floor or office at any given time. This is done in full GDPR compliance by literally counting heads without looking at faces. In the short term, the information can be used to save energy, for example, says Ubiqisense's Head of Software Klaus Grouleff:
"If data can show that most people arrive at 9 AM, the ventilation doesn't need to run from 7 AM. In the longer term, the solution can make it easier to better organise the square metreage, he emphasises. The number and size of meeting rooms and offices can be customised if most meetings count 2 - 4 people and many people work from home. This enables facility managers to optimise their buildings and save both CO2 and money.
Ubiqisense had an ambition to make the solution both faster and more accurate, says Klaus Grouleff. "We had some different ideas and had started to explore the area of convolutional neural networks, but we ran into some challenges."
Previously, Ubiqisense had modified a ready-made software package and customised the models to their own needs, adds his colleague, Development Specialist Kim Andersen. "But now there were no more parameters we could adjust." To improve the solution further, they needed to take a new approach, and that's what the AI Denmark programme was designed to help with.
In the initial conversations, three of the Alexandra Institute's experts independently recommended switching to a machine learning framework better suited for computer vision. In the AI Denmark programme, the institute did a broad search of architectures to find the strongest candidate, says Kim Andersen.
Ubiqisense's challenge is rather unique: To solve the task, the precision must be high so that you know exactly how many people are exactly where. At the same time, it is both an ethical and legal requirement that the information coming from the sensor is anonymised and that you don't know who is in the room.
This particular challenge had made it difficult for Ubiqisense to find the right solution, says Kim Andersen: "There are solutions that can find all sorts of very distinctive objects in images - from elephants and tigers to cakes." However, this is quite different from counting people anonymously and with high precision, which is why the Alexandra Institute helped Ubiqisense identify the best way to solve the task.
During the AI Denmark project, Ubiqisense clarified what they needed to do to increase the speed and accuracy of their sensors. In a subsequent short programme with the Alexandra Institute, the company implemented the recommendations with excellent results, says Kim Andersen. "We have halved the number of errors and thus doubled our precision," he states. "And we've done a lot of research and saved time compared to if we had to do it ourselves." He believes that it has made a big difference for Ubiqisense to gain knowledge and tools that enable them to continue developing on their own.
For Kim Andersen, the AI Denmark programme has created great value in terms of acquiring new knowledge. "What I value most is that I learnt a lot," he emphasises. "It was a unique opportunity to bring back knowledge instead of just hiring a consultancy firm, getting the job done and then being at a loss again afterwards," he says.
According to him, smaller companies in particular can gain a lot from an AI Denmark programme. "It's a great opportunity in a company with a limited number of people to spar with," he points out. He appreciated the Alexandra Institute as a sparring partner from whom he learnt a lot, partly because the institute has many different experts.
The new knowledge is already being incorporated into Ubiqisense's product, says Klaus Grouleff. He is looking forward to customers benefiting from more valid data and better analyses, making it easier for them to make decisions about their buildings in the future. After the COVID-19 lockdown, there is an increasing focus on making better use of square metres, he says.
"We returned to workplaces that were completely changed: Some people work at home and others come to the office, but you don't really know when."
"With our solution, you get data so you can see how your building is being used real-time. And with that knowledge, Ubiqisense's customers can make better use of their buildings and save on costs, but also reduce the consumption of energy and other resources. This benefits both the climate and the environment. According to the International Energy Agency, the operation of buildings accounts for 27% of CO2 globally, making it a key area to optimise," he explains.
"For example, across our largest customers, we see an average occupancy rate of 34% in meeting rooms. This means that two-thirds of the meeting room capacity is not being utilised," he continues. "Meanwhile, the rooms' ventilation, heating and electricity are running at full capacity. The potential to reduce energy and emissions from buildings is huge and we need to tap into that, and regardless of whether people have fixed seats, it can be done in a GDPR-secure and privacy-preserving way," he concludes.
The article is authored by Alexandra Instituttet.
