The DECENTER project utilizes Leti IP, aimed at Edge AI
LONDON — A joint project funded by European and South Korean research programs aims to develop an IoT platform empowering emerging artificial intelligence (AI) applications with on-demand computing at the edge of networks.
As part of the project, research institute Leti announced that its sensiNact IoT middleware will be the core of the platform under development in the three-year development program.
The joint project, called DECENTER, is focused on integrating the IoT, AI, the cloud, edge, fog computing and smart contracts, together with a secure blockchain. It aims to develop a platform facilitating an ecosystem in which computing and IoT resources (processing, memory, storage, connectivity, sensing, actuating) are orchestrated in multi-cloud, federated environments. In this “fog computing platform,” it envisages that all providers can share resources and be rewarded through the automatic execution of smart contracts logged and monitored via blockchain-based technologies.
The federated fog platform intends to address processing of the ever-increasing amount of data continuously gathered from a myriad of heterogeneous IoT sensors and appliances. In the joint EU-Korea project, this will be demonstrated in various scenarios, such as smart homes, smart cities, smart construction sites and robotics.
Leti’s IoT middleware, sensiNact, is a unified framework for integrating and managing IoT devices via generic application programming interfaces (APIs). It collects, aggregates, and securely scripts data from a wide range of communicating objects, regardless of the network communication protocol. As a key component of the DECENTER platform, it will provide distributed sensing and actuation resources that will be shared and used by the blockchain-based resource orchestrator to ensure the accounting and the integrity of the transactions by smart contracts.
The DECENTER provides reactive and effective computation to deploy (distributed) AI applications. It is a decentralized ecosystem in which resources owned by multiple providers can be harmoniously orchestrated throughout the cloud-to-things continuum (vertical) in dynamically created multi-cloud environments (horizontal).
In a telephone interview with EE Times, Levent Gurgen, Leti’s senior expert in IoT, said that a key outcome is to enable machine learning in edge devices with AI algoritms being deployed in a distributed manner. “This will help us reach the goal of doing as much as possible at the edge and address latency and privacy issues," Gurgen said.
He said that Leti has been working on its sensiNact IoT middleware since 2010, and the project was extended to open-source under the Eclipse Foundation. The core platform coordinates data from multiple devices and sensors, and its sensiNact Studio enables creation of applications, deployment to the platform, and lifecycle management. It has already been deployed in several collaborative projects, such as the EU-Japan ClouT project, whereby sensiNact enables access to over 500,000 physical and virtual devices across European and Japanese cities (Santander, Genova, Fujisawa, and Mitaka).
Decentralization from the cloud to the edge is a challenge for AI technologies applied to large heterogeneous systems. Another issue is how to ensure timely and effective responses for critical applications such as braking. While processing compute- and storage-intensive resources will still be hosted in the cloud, the argument that many companies are increasingly making is that it is necessary to now have timely and effective computing at the edge that provides support. Leti said that this can help significantly in scenarios such as improving safety at pedestrian crossings or for workers and assets in logistic and construction areas.
Gurgen said that at the end of three years, Leti hopes that the platform will be implemented in products developed by some of its industrial partners — such as SK Telecom. Leti also hopes to spin-off its sensiNact-related activities.
The sensiNact middleware supports various IoT protocols (ZigBee, CoAP, EnOcean, LoRa, Sigfox, MQTT, XMPP, etc.); provides access on-demand, periodically or event-based to real-time data for online analysis; enables rapid creation of new bridges to emerging protocols and dynamically integrates them to the running platform; and creates automation applications exploiting the data and action sources, deploys them to the platform, and manages their lifecycles.
The European-Korean consortium combines complementary expertise — the European partners with their specialism in IoT and blockchain technologies and the South Korean partners with their expertise in AI and cloud computing. European partners in the 11-member project are Fondazione Bruno Kessler (Italy), Atos (Spain), CEA-Leti (France), Comune di Trento (Italy), Robotnik (Spain), and University of Ljubljana (Slovenia). South Korean partners are Korea Electronics Technology Institute, Gluesys, Daliworks, LGU+, and Seoul National University.