Demonstration 1: Deep Learning-Based Communication Over the Air; The world’s first communications system entirely based on neural networks

Presenter: Dr. Jakob HOYDIS (

Description: We have introduced a new way of thinking about communications as an end-to-end reconstruction optimization task using autoencoders to jointly learn transmitter and receiver implementations without prior knowledge. This approach is applicable to any type of channel and shown to achieve competitive performance with respect to state-of-the-art schemes on real hardware. We believe that this is just the beginning of a much more comprehensive study into applications of deep learning for the physical layer. We are truly excited about the possibilities it could lend towards future communications systems as the field matures.

Demonstration 2: Multi-Armed bandit Learning in Iot Networks (MALIN)

Presenters: Rémi Bonnefoi (1), Lilian Besson (1)(2) and Christophe Moy (3)
(1) CentraleSupélec/IETR , 35576, Cesson-Sévigné Cedex, France
(2) Univ Lille, CNRS, INRIA, SequeL Team, UMR 9189 – CRIStAL, F-59000 Lille, France
(3) Univ Rennes, CNRS, IETR – UMR 6164, F-35000, Rennes, France

Description: In this demonstrator we use Multi-Armed Bandit (MAB) Algorithms for the purpose of channel selection in IoT networks. These algoritms allow IoT devices avoiding overcrowded channels, i.e., with a high Packet Loss Ratio (PLR). In the proposed scenario, the learning is operated based on the acknowledgement transmitted by the base station. As a consequence, it does not require any cooperation between devices and consequently does not require any extra signaling.

Demonstration 3: Multi-source energy management for Long Range IoT nodes

Presenters: Philip-Dylan Gleonec, Wi6labs, Univ Rennes/IRISA

Description: Energy harvesters are used in Internet of Things (IoT) networks to provide more energy to the nodes and enhance their autonomy. However, the nodes must adapt their consumed energy to variable energy harvesting conditions. Typically, an energy budget estimation (EBE) algorithm is used to calculate an energy budget based on the current energy capabilities. In this demonstration, we will show the implementation and a comparison of multiple EBE algorithms on a real world LoRaWAN platform powered by different energy sources (indoor light, thermal, …). All nodes will be connected to a remote LoRaWAN application server through a LoRa gateway.

Demonstration 4: Comparative Performance Analysis of Mobile Network Operators

Presenters: Alper Senyildiz, Türk Telekom Labs

Description: Mobile Network Operators (MNOs) are interested to know how their competitors’ performances vary at certain region and in certain day of the month for each cellular network technology (CNT) in order to invest intelligently in future. Therefore, network performance comparisons of the key parameter indicators (KPIs) between MNOs is of interest. In this demo, we demonstrate and visualize performance comparisons of MNOs for assessing the relative Key Parameter Indicators (KPIs) differences in large geographical regions. This demonstration can provide valuable insights to MNOs in order to evaluate performance differences for some important KPIs.

Demonstration 5: Partial reconfiguration on ARM/FPGA Platform for Vertical Handover for WIFI/Mimax

Presenters: Mohamad-Al-Fadl Rihani, Lebanese University, Faculty of Engineering III, Beirut, Lebanon and IETR, Jean-Christophe Prévotet,  Fabienne Nouvel, IETR-INSA, Rennes France

Description: In wireless networks, end-nodes are able to detect the presence of multiple standards and to switch between them is they are available on chip. In this context, it becomes interesting to design an on-line reconfigurable communication system thanks to a Vertical Handover Algorithm (VHA) that allows selecting the best wireless standard. In this demo, we propose to implement this mechanism using the Partial Reconfiguration (PR) technique on a SoC platform (ARM-FPGA). We apply VHA between WIFI and Wimax. The demo simulates the mobility of an end-node in a WIFI-WiMax network on a GUI Interface connected to the ZedBoard. On the SoC, the VHA senses specific parameters and decides accordingly to reconfigure a unified PHY layer before applying partial reconfiguration on the device.

More demonstrators will be listed soon