Initial situation and problem definition:

With the breakthrough of Deep Learning (DL) in image and speech processing and the multiple victories of Artificial Intelligence (AI) in games such as Chess or Go, the AI winter has ended and the technology has once again become the focus of scientific and public interest. On the other hand, there are high demands on future communication systems, in which more and more devices are connected with each other, thus requiring ever increasing efficiency and performance.

Therefore it is worthwhile to investigate and transfer AI for its applicability in communication technology. For this reason, the ANT has decided to include AI in its main research topics and to investigate how current communication systems can benefit from AI algorithms.

Objective:

In this project, fundamental concepts of AI procedures are investigated and evaluated for their applicability in telecommunication systems. Thereby concepts for the physical layer are in the foreground

As an example of how AI can be useful for communication systems, we use two recent publications of the ANT, which are shown below. By using model-driven neural networks, we are able to achieve performance improvements as well as lower computational complexity for the equalization of large multiple-input multiple-output (MIMO) systems compared to state-of-the-art approaches.

Details