Smarter applications with edge intelligence

We’ve had our head in the cloud for some years, but now there’s a shift towards living on the edge. Bringing computing power closer to data sources will allow for smarter applications.

Drone Photo: Mostphotos
Photo: Mostphotos.

On April 4, 2023, the Department of Computer and Systems Sciences hosted a Tech Tuesday event for the tech community in Kista. Professor Javid Taheri was the speaker, sharing interesting insights from his research on edge intelligence.

“This is a topic that I hold close to my heart. Edge intelligence is the beautiful baby from the marriage between AI and machine learning”, said Javid Taheri.

IoT is here to stay

But to understand what edge intelligence is, we have to start by talking about Internet of Things (IoT).

“IoT is present in every aspect of our lives. It affects how we travel, how we live at home, how we work in factories… And IoT is here to stay. The forecast is that the IoT market will continue to grow.”

All everyday items that have an IP address fall into the category of IoT devices. Watches, refrigerators, vacuum cleaners and cars are just a few examples.

Javid Taheri presenting at Tech Tuesday April 4, 2023.
Javid Taheri. Photo: Åse Karlén.

“These devices can be producers or consumers of data – they can be senders or receivers”, said Taheri.

Cloud vs edge

Together they collect huge amounts of data, and they are usually connected to big data centers – the cloud. According to Javid Taheri, IoT creates and processes some 75 percent of all data, but 90 percent of the data is never put to use.

“Think of a steel oven in big factory where the temperature is measured every second. What should they do with all that data? To be able to consume data, we need to bring the computation part closer to the data”, said Javid Taheri.

And this is where edge computing comes in. Action points are closer to data which means, for example, that the oven in the example above will be able to turn off when it gets too hot.

Artificial intelligence (AI) is not needed for these kinds of manoeuvres, Taheri explained. But AI can help edge – and edge can help AI.

“One of difficulties with AI is that you need to train it on a lot of data. Edge can take massive amounts of data, zip it and then send it to AI.”

AI can be helpful when harder decisions need to be made, for example if a robot is misbehaving and must to be corrected.

“AI will be more useful when it’s closer to data”, concluded Javid Taheri.

Edgified applications

He mentioned some of the challenges that edge intelligence is facing, associated with execution, networking, security and software.

The big opportunity is that edge intelligence makes applications smarter, whether we’re talking about drones, robots, surveillance systems or Industry 4.0.

“Large companies like Ericsson, Huawei and Vodaphone are building edge computing platforms. Smaller companies can use these infrastructures to provide services, and ‘edgify’ their applications.”

“They can, for instance, build applications that let cars talk to each other. Another example is applications for drones to fly under bridges, take photos and process them to see if maintenance is needed”, said Javid Taheri.

About the event
Tech Tuesday is a monthly networking and knowledge-sharing event for the tech community (private companies, academic institutions and other organisations) in Kista. It is organized by Kista Science City.

On April 4, 2023, the Department of Computer and Systems Sciences (DSV) at Stockholm University was the host of the event.

Professor Javid Taheri, Karlstad University and Ericsson, was the speaker. The topic was “Edge intelligence: from theory to practice”.

For more information about DSV’s research in this area, contact professor Rahim Rahmani or professor Panagiotis Papapetrou.

Text: Åse Karlén