AI@EDGE: the artificial intelligence of the future - safer, faster, more open

AI - driving the new industrial revolution in Europe

To transition from artificial intelligence (AI) centralised on the cloud to embedded AI, getting closer to end users: this is the primary objective for the 19 European partners who have come together to form AIAEDGE. Coordinated from Italy by the Bruno Kessler Foundation, the project, which was launched in January, will receive more than eight million euros’ worth of funding over a period of three years as part of the Horizon 2020 call for proposals ICT52 2020. It was the only project selected from the thirteen submitted by Inria project teams to this call for projects, the aim of which was to support the development of information and communication technologies (ICT). There was a particular focus on artificial intelligence, a subject given priority status by the upcoming research and innovation programme “Digital Europe”.

For Inria, the total amount of EU funding stands at more than €300,000. This will be used to recruit a PhD student and an engineer, who will join the two project leaders: Jérôme François, a researcher at Inria, and Abdelkader Lahmadi, a teacher-researcher at the University of Lorraine, both from the RESIST project team. They will work together alongside figures from the worlds of academia, science and industry, all motivated by one common goal: To develop “a secure and reusable artificial intelligence platform for edge computing beyond 5G networks.” The project promises a whole host of potential applications in manufacturing, telecommunications, health, logistics and transport.

Moving towards edge AI

“Increasingly common in the deployment of services via the internet, AI is no less compartmentalised: it is stored either on connected devices (smartphones, tablets, connected objects, etc.) or on the cloud”, explains Jérôme François. “With network infrastructures such as 5G, for example, it might be possible to deliver services end-to-end with more bandwidth, connectivity, security, etc.” This is where the concept of “edge” comes into edge computing. Unlike with the cloud or data centres, which keep data in the one place, this type of emerging IT architecture can be used to process data on the edges of a network, getting closer to users. It does this through devices which generate data, such as connected objects, smartphones, computers or local servers.

To develop robust and reusable models

Whether it’s reducing data processing latency, improving processing and analysis capacities or optimising the effective bandwidth, combining artificial intelligence with edge computing could have a range of benefits. But there are also challenges involved in implementing it: “What we want to do is to divide tasks up between connected devices, servers and networks and to ensure they interact smoothly with each other over time, but that’s not all”, explains Jérôme François. “It will be necessary to deploy applications and develop reusable and robust AI models in order to manage different types of resources while keeping them secure, particularly against the threat of reverse engineering. This is made more complicated by the increase in the volume of resources being shared between users.”

From driverless vehicles to infotainment on-board aircraft

Issues surrounding heterogeneity, security and data volumetrics are central to the use cases that AIAEDGE will be focusing on. Of the four, one concerns collaborative perception for vehicle networks: “This involves driverless vehicles sharing data in order to predict and respond better to emergency situations such as accidents or traffic jams”, explains the researcher.

Another case RESIST is taking a particular interest in is secure and multipartite AI for the internet of things, particularly in the industrial sector.  “Whether it’s production lines, cameras or temperature sensors, the challenge lies in grouping together different data sets from different systems in different places and ensuring they remain secure”, explains Jérôme François. AIAEDGE will also explore the use of drones for monitoring infrastructure and infotainment on-board aircraft, looking “at ways of developing more personalised content and services that will follow passengers on their journeys”.