Bielefeld University

The multitude of variously communicating everyday devices is often referred to as the Internet of Things (IoT for short). The range of applications for IoT devices extends from mobile devices to self-driving cars and industrial robotics. However, 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 implement quickly and efficiently by conventional means.

Twelve partners from four European countries - Germany, Poland, Portugal and Sweden - as well as Switzerland, are working together in VEDLIoT. The international team of researchers is using machine learning methods, including Deep Learning, to implement the necessary functions. "In Deep Learning, the underlying network has intermediate neuron layers in addition to input and output layers. This realizes a kind of abstraction that enables complex system behavior," says Jens Hagemeyer, a member of the Cognitronics and Sensors research group and technical coordinator of the project. "We provide the information; the machines learn and decide for themselves."

With the autonomous learning of the VEDLIoT platform, IoT devices are expected to 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 and computing architectures to be flexibly combined to create a universal platform. In addition to the appropriate selection of micro servers, a flexible, high-performance communication infrastructure between the different micro servers is essential for an efficient and fast realization of the application. For this purpose, the VEDLIoT hardware platform combines various novel communication infrastructures that enable both high-performance and low-latency communication between the micro servers.

In the VEDLIoT project, the smart mirror serves as a demonstrator for the class of smart home applications. It represents the interface between the occupants and the smart environment and consists of a combination of a mirror and a display. It visualizes personalized information (e.g., messages or the status of the smart-home environment) and features face, object, gesture and voice recognition. Instead of performing the necessary calculations in the cloud, they are performed locally on the energy-efficient VEDLIoT hardware platform. In addition to the energy-efficient computation, the low-latency communication infrastructure will also be shown as part of a live demonstration.