Background and Significance

During the last century, the telecommunications industry was devoted to improving human connectivity, targeting seamless worldwide coverage to cope with increasing data rate demands and mobility requirements. Over the past decade, a major shift towards machine-to-machine communication—often referred to as the Internet of Things (IoT)—has occurred. Due to the fact that there are orders of magnitude more machines that humans, this has lead to a range of new design challenges. The first key challenge is simply the scale of the network, where billions of devices must be connected with all the associated coordination difficulties. The second key challenge is the character of the data to be transmitted. Unlike most human communications, many machines transmit relatively small quantities of data with significant variations in quality and latency constraints. The third key challenge is the fact that communication is increasingly integrated with how machines perform sensing and are used as actuators in networks of controllers. An important consequence is that how communication systems are designed can no longer be treated independently of how the machines learn to interact with their environment, both in terms of system identification and control.

In MARACAS, we address these challenges from a communication theory perspective, drawing on tools from statistical signal processing, information theory, machine learning as well as experimental campaigns. In particular, we are responsible for the FIT/CorteXlab platform for reproducible experiments in wireless communications. We also contribute the development of 5G systems and beyond as well as emerging applications of communications in Cyber Physical Systems.

The research in MARACAS is carried out in four axes:

Axis 1: Fundamental limits of reliable communication systems. In this axis, we seek to understand what communication systems are capable, in terms of quantities such as data rates, error probability, energy efficiency, required computational resources and latency. By exploiting tools from information theory, statistics, game theory and optimisation theory, we aim to characterise the optimal performance in order to guide algorithm design.

Axis 2: Algorithms and protocols. In this axis, we develop new algorithms and protocols for 5G communication systems and beyond. The focus is typically on PHY-layer issues, such as channel estimation, resource allocation, decoding algorithms. With the need for integrated communication, sensing and control, we are also investigating codesigned algorithms. With the recent advances in machine learning, we are also applying these tools to communication system design.

Axis 3: Experimental validation. In this axis, we develop experimental platforms such as the FIT/CorteXlab platform, and carry out experimental campaigns to investigate the real-world performance of our algorithms. Part of this work involves establishing data sets, which are made publicly available and are critical for advances in machine learning in wireless communications.

Axis 4: Emerging application areas. As communication theory can play a key role beyond the wireless networks in the IoT, we also investigate other settings. These include molecular communications, smart grid, and the use of communication theory to design Cyber Physical Systems.

Recent Progress

Axis 1: Our recent work on fundamental limits of reliable communication has focused on two main themes: communication in the IoT; and sensitivity analysis of the information capacity. To design IoT communication systems, it is critical to understand the impact of low latency constraints on the probability of error and achievable data rates. Our work in this area has established new characterisations of the data rate/error probability tradeoff in the finite block length regime [1]. We have also investigated the impact of very large numbers of transmitting devices, establishing new characterisations of the resulting interference statistics [2]. In particular, we have shown that multivariate alpha-stable models naturally arise and established information theoretic limits on communication with additive alpha-stable noise [3].

A key question in information theory is the impact of model parameters on quantities such as the channel capacity. While this has been characterised in settings such as additive Gaussian noise channels, for non-Gaussian noise models much less is known. By exploiting techniques from parametric optimisation theory, we have established new bounds for changes to the channel capacity as the noise distribution or input constraints are varied [4,5]. This work also provides insight into the structure of the optimal input distribution [6].

Axis 2: The optimization of IoT access techniques was the objective of the ANR Ephyl collaborative project, where we studied different solutions at the PHY and MAC layers as presented in [8]. The main question Maracas group addressed in this research is the detection of simultaneous random transmissions from distributed nodes. The underlying mechanism is a coded slotted Aloha allowing to avoid hand-skake mechanisms. Each node can transmit randomly and the receiver tries to detect several packets simultaneously. Our objective is to identify a good code family, and to determine the fundamental trade-off in terms of nodes density versus reliability. During this year, we focused on the detection of a small subset of simultaneous active nodes, exploiting optimal detection. We developed a MAP based iterative detector at a multi-antennas receiver in [9]. We also proposed a low complexity detector in [10].

Following the artificial intelligence tsunami, the research community in wireless systems (both industry and academia) is engaged in a strong competition to determine how this revolution could change the paradigm of wireless networks. Following preliminary studies, we investigate the potential of deep learning in radio communications. The central question is to identify which processing could take advantage from neural networks against classical approaches. Our joint strategy with Nokia follows: we target the production of a huge set of experimental data with FIT/CorteXlab to facilitate the comparison of different solutions and to train neural networks on real data. We currently investigate three original problems : transmitter identification from its RF signature [11], self-synchronization procedures based on neural networks, and dirty RF compensation.

Interference management and resource management is a very complex problem in wireless environment. The capacity region is known for some specific scenarios and some specific channel conditions. But the optimal performance relies on perfect feedback mechanisms, to get channel state information at the transmitters and to coordinate them. Topological interference management (TIM) is a seducing framework to balance performance with feedback complexity. In the context of TIM, We have developed new algorithms to allow partial coordination between interfering transmitters [12], relying only on some partial interference information. This approach suits particularly well with the requirements of PMR networks, since their deployments is not optimized. The algorithm relies on an association of degrees of freedom evaluation, graph theory and interference alignment.

Axis 3: CorteXlab relies on Minus, an experiment conducting middleware which allows users to submit experimental tasks to the platform, handles the automatic execution of these experiments, and gathers their results. The initial design for Minus relies on a fixed toolchain (mainly composed of GNURadio, hardware drivers, and additional external or in-house software or GNURadio blocks, FPGA tools, etc.). Experimenters are supposed to use this fixed toolchain in a batch-like workflow. It is hard for experimenters to extend the limits of the fixed toolchain (e.g. to use a custom library or software, or a different version of GNURadio), and the development phase of an experiment can be painful due to the batch-like interface. To improve this, we have developed a new experimental workflow based on docker images and containers which allows experimenters to use our in-house provided docker images, adapt them if needed, or even create completely custom ones. These images have the benefit that they can be used identically on the experimenters’ workstations, on the CorteXlab platform, or another platform, and they can be used interactively if needed, even on CorteXlab. This increases greatly the ease of use of the platform, the reproducibility and share-ability of experiments, and the breadth of its usage.

In this work we developed an experimental setup for dense IoT access evaluation using the FIT/CorteXlab radio testbed. The aim of this work is to provide a customizable and open source design for IoT networks prototyping in a massive multi-user, synchronized and reproducible environment thanks to the hardware and software capabilities of the testbed.
The massive access feature is managed by emulating a base station and several sensors per radio nodes. Two categories of modular network components are used in our design: a base station unit and a multi-sensor emulator unit. These components are separately hosted in dedicated and remotely accessible radio nodes.

Axis 4: Our recent work on emerging communication systems has focused on the nanoscale, where devices (e.g., synthetic cells) communicate by emitting small quantities of molecules. Here, information can be encoded in the number, type, and timing of the released molecules. We have studied in particular constraints imposed on communication by the presence of nearby biochemical systems (e.g., natural cells), where it is necessary to ensure that natural behavior is not changed. To this end, we have established information theoretic limits on the number of messages that can be transmitted [7]. This work also intimately links recent advances in communication with security constraints to the design of molecular communication systems.

Selected Publications

[1] Philippe Mary, Jean-Marie Gorce, Ayse Unsal and H. Vincent Poor, “Finite block length information theory: what is the practical impact on wireless communications,” Proc. IEEE GLOBECOM Workshops, 2016.

[2] Malcolm Egan, Laurent Clavier, Ce Zheng, Mauro de Freitas and Jean-Marie Gorce, “Dynamic interference for uplink SCMA in large-scale wireless networks without coordination,” EURASIP Journal on Wireless Communications and Networking, Aug. 2018.

[3] Mauro de Freitas, Malcolm Egan, Laurent Clavier, Alban Goupil, Gareth W. Peters and Nourddine, Azzaoui, “Capacity bounds for additive symmetric alpha-stable noise channels,” IEEE Transactions on Information Theory, vol. 63, no. 8, pp. 5115-5123, Aug. 2017.

[4] Malcolm Egan, Samir Perlaza and Vyacheslav Kungurtsev, “Capacity sensitivity in non-Gaussian noise channels,” Proc. IEEE International Symposium on Information Theory, 2017.

[5] Malcolm Egan and Samir Perlaza, “Capacity approximation of continuous channels by discrete inputs,” Proc. Annual Conference on Information Sciences and Systems, 2018.

[6] Alex Dytso, Malcolm Egan, Samir Perlaza, Shlomo Shamai (Shitz) and H. Vincent Poor, “Optimal inputs for some classes of degraded wiretap channels,” Proc. IEEE Information Theory Workshop, 2018.

[7] Malcolm Egan, Valeria Loscri, Trung Q. Duong and Marco Di Renzo, “Strategies for coexistence in molecular communications,” IEEE Transactions on NanoBioscience, vol. 18, no. 1, pp. 51-60, 2019.

[8] Guillaume Vivier et al., “Beyond LoRa and NB-IoT: proposals for future LPWA systems,” Proc. EuCNC, 2019.

[9] Diane Duchemin, Lelio Chetot, Jean-Marie Gorce and Claire Goursaud, “Coded random access for massive MTC under statistical channel knowledge,” Proc. SPAWC, 2019.

[10] Diane Duchemin, Lelio Chetot, Jean-Marie Gorce and Claire Goursaud, “Détecteur pour l’accès aléatoire massif entre machines avec connaissance statistique du canal en lien ascendant,” Proc. GRETSI, 2019.

[11] Cyrille Morin, Leonardo Cardoso, Jakob Hoydis, Jean-Marie Gorce and Thibaud Vial, “Transmitter Classification With Supervised Deep Learning,” Proc. CROWNCOM, 2019.

[12] Hassam Kallam, Leonardo Cardoso and Jean-Marie Gorce, “Topological Interference Management: Trade-off Between DoF and SIR for Cellular Systems,” Proc. ICT, 2019.

Comments are closed.