VEDLIoT
VEDLIoT
Very Efficient Deep Learning in Internet of Things
EU project VEDLIoT aims to develop a next-generation IoT platform
Autonomous vehicles and smart home devices are becoming increasingly complex. A new machine learning-based system is being developed to make the software and hardware used for these applications more robust, powerful and energy-efficient. The new VEDLIoT project is funded by the European Commission with around eight million euros over a period of three years. The project is coordinated by the CoR-Lab at Bielefeld University.
In a smart home - a "smarthome" - residents have devices at their disposal to make their lives easier: Imagine a refrigerator that reorders food when it runs low and can communicate with the oven at the same time. Such devices and modules are part of the Internet of Things (IoT for short). IoT devices are connected to a network where they record, store, process and transmit data. Applications for IoT devices also include self-driving cars and industrial robotics.
The goal is to develop an energy-efficient hardware platform and tool chain for ML-based IoT applications. From left: Jens Hagemeyer, Dr. Carola Haumann and Prof. Dr.-Ing. Ulrich Rückert from Bielefeld University.
"Computer and IoT systems are becoming increasingly efficient. This enables us to solve more challenging problems and accelerate automation to improve our quality of life," explains Professor Dr.-Ing. Ulrich Rückert, who heads the Cognitronics and Sensor Systems research group at Bielefeld University as coordinator of the new VEDLIoT project. "But the volume of data that is collected and processed is enormous - and the computing power required for this is very high. In addition, the algorithms are often too complex to generate solutions quickly and in a reasonable amount of time."
Artificial intelligence versus conventional methods
Twelve partners from four European countries - Germany, Poland, Portugal and Sweden - as well as Switzerland, an EU associate country, are working together on the project. Instead of relying on conventional methods, such as those from statistics, the international team of researchers is using machine learning methods, including Deep Learning, which employs artificial neural networks. "In Deep Learning, the underlying network has intermediate neuron layers in addition to input and output layers. This allows a kind of abstraction to be realized, which thereby enables complex system behavior," says Jens Hagemeyer, an electrical engineer who is a member of the Cognitronics and Sensor Systems research group and is also the technical lead on this project. "We provide the information; the machines learn and decide for themselves."
The VEDLIoT platform's autonomous learning is expected to help IoT devices achieve higher performance while becoming more energy efficient. To this end, the project is developing a modular hardware platform that allows micro-servers of different performance classes to be combined on a flexible carrier. "Depending on the requirements of the application, the servers can be individually configured on the carrier, creating a universally applicable platform," Hagemeyer said. System failures are also prevented with the new system: "If a server fails, for example due to a weak radio network, the entire device can continue to operate. In the best case, the user of a self-driving car would not even notice the server failure."
Open call for additional project partners
"Some of the project partners have already been working together for many years," said Dr. Carola Haumann, project manager and CoR-Lab's deputy managing director. The project includes seven universities and research institutes working in the field of artificial intelligence and the Internet of Things. The other project partners are companies of various sizes, from start-up EmbeDL to multinational Siemens.
There is still time for more companies to join the project: "We expect to fund at least ten more use cases in this project - in addition to the existing applications in the automotive, automation and smarthomes sectors. That's why we want to involve more companies," explains Haumann. A prototype of the platform is expected to be ready for use by mid-2022. "The results from these different applications will be incorporated into the IoT platform throughout the project," says Jens Hagemeyer. "This will enable us to continuously improve the platform." The project was launched in November 2020, and an in-depth workshop with all project partners is planned for early December.
The project is expected to be completed by the end of 2023. VEDLIoT is funded by the Informatics and Communication Technologies funding line of the EU's Horizon 2020 research framework program. The name VEDLIoT is an acronym for "Very Efficient Deep Learning in IoT."
Researchers on the new VEDLIoT project are developing a modular hardware platform that could be used in a range of applications, from a smart mirror to smarthome devices.
In addition to Bielefeld University, research institutions and universities are also involved in the project: Chalmers University of Technology in Gothenburg (Sweden), University of Neuchâtel (Switzerland), University of Osnabrück (Germany), University of Gothenburg (Sweden), Swedish Research Institutes (RISE) in Gothenburg (Sweden), and FCiências.ID, a research and development association in Lisbon (Portugal).
Participating companies are: Antmicro in Poznan (Poland), EmbeDL in Gothenburg (Sweden), Siemens based in Munich and Berlin (Germany), Christmann Informationstechnik + Medien in Ilsede (Germany) and Veoneer in Stockholm (Sweden).
More information:
Contact:
Prof. Dr.-Ing. Ulrich Rückert, Bielefeld University
Faculty of Technology / CoR-Lab
E-mail: rueckert@techfak.uni-bielefeld.de