The AI4CODE project brings together 6 research team with strong expertise in the design, decoding and standardization of forward-error-correction codes. The aim is to develop skills in artificial intelligence and machine learning, and to explore how learning techniques can contribute to the improvement of code design methods (by using less parameters, more relevant heuristics, producing stronger codes) and decoders (better performance, reduced complexity or energy consumption), on selected scenarios of practical interest for which a full theoretical understanding is still lacking. The proposed methodology is to augment legacy design methods and decoders with learning capabilities or decision support systems wherever relevant, rather than replacing them by a generic, black-box neural network, so that we can inspect the trained solutions and try to infer why they work better. Our ultimate goal is to obtain new theoretical hindsight that could translate into better codes and decoders.

Optical fiber constitutes the backbone of the modern information networks. Long range communications are based on coherent detection and spectrally efficient digital signal processing. Future optical communications are expected to involved radically new optical links to support short range data center interconnections with cost constraints. In both cases, advanced high-dimensional modulation and fine matching to the physical channel are needed to significantly improve the spectral efficiency of the system. The MUSICO project aims at providing modern DSP solutions for the next generation of ultra-efficient and agile optical networks by focusing on high-dimensional signaling in combination with nonbinary joint system design. The key impact of the project is that it will revisit completely the joint signaling schemes for optical communication and will address them from a viewpoint of a new, more accurate and practical, optical channel model. The main outcome of the project will therefore be an entirely new set of design tools for joint modulation and coding, adapted to the multi-dimensional and potentially non-linear nature (e.g., XPol model) of the optical channel. First, we expect to double (or triple) the spectral efficiency in certain operational scenarios as it is observed when PDL occurs in long-haul transmissions. Signaling, coding, and channel learning should be carefully addressed, and they lay at the core of this project. Second, for short range communications, we expect to be able to propose schemes that permit novel and high-efficient N-PAM signaling to be deployed on a new generation of low-cost products based on direct detection.

We study communication problems where undetectability or location privacy is required. Such secrecy requirements are important, e.g., for “smart devices” which only communicate scarcely to send short messages, where much sensitive information is contained not in the messages, but in who is sending the message, when, and from where, etc. Recent works on “covert communication” provided a theoretical framework for analyzing undetectability. For a broad class of channels, the “square-root law” holds: the number of bits that can be communicated covertly is proportional to the square root of total communication time. This implies that the maximum bits per second for covert communication tends to zero as total communication time grows. In this project we explore new channel models and try to find scenarios where positive-rate covert communication is possible. We also further explore the square-root scenario to bridge the gap between theoretical results and real-life applications.

The SoCoRe! project (Open science for Conservation / Restoration of the cultural heritage) aims at structuring and opening data produced by the activities of conservation / restoration of the tangible cultural heritage. Supported by the Fondation des Sciences du Patrimoine, which gathers a very large scientific community in this field, the SoCoRe! project continues the effort, led by the foundation, of designing common models, vocabularies, platforms and tools for sharing conservation / restoration data. The goal of the SoCoRe! project is to build the tools for producing data that respects the common models, for storing, sharing and querying data produced by the community, and to specify the good practices and the procedures for implementing this approach for an open science.

The publication of 3GPP Release 15 in June 2018 paved the way for the new 5G air interface, making a new step towards the new generation of mobile networks. The work on the development of a new radio interface dedicated to Internet of Things (IoT) connectivity should then start in 2020. It is expected to replace or to complete the Narrow Band-IoT and LTE-M interfaces originally specified in Release 13 in order to achieve the multi-service capability of 5G. Actually, IoT use cases can be segmented into two categories: “critical” applications (traffic safety, automated vehicles …) and “massive” applications (smart buildings, transport logistics …). Massive applications are characterized by an expected high density of connected devices (1 Million/km² from IMT-2020 requirements), small data payloads, as well as stringent constraints on the device energy consumption and cost. Maximizing the spectral efficiency of an IoT network is a key pre-requisite for providing massive connectivity. At link level, it can take advantage of powerful error control codes such as Non-Binary (NB) codes. At system level, reducing “meta-data” throughput, i.e., exchange of information linked to signaling, synchronization and identification is the new paradigm of massive IoT network. This requires encompassing those functions in a single well protected frame. The first wave of IoT standards are far from achieving the reliability and spectral efficiency targets: they implement sub-optimal Forward Error Correction (FEC) schemes such as convolutional or Turbo codes combined with repetition codes (EC-GSM, Narrow Band-IoT and LTE-M [3GPP]), simple Hamming codes (LoRa), or simply omit any FEC capability (SigFox). The aim of the QCSP project is to contribute to the evolution of IoT networks by defining, implementing and testing a new coded modulation scheme dedicated to IoT networks. The “big bet” of the project is to work on the emergence of NB codes combined with a Cyclic Code Shift Keying (CCSK) modulation. This new coded modulation scheme, called CCSK-NB-code [ABA13, ABA14], can be easily implemented in a cost efficient way at the device side. CCSK-NB-code provides several advantages compared to state of the art waveforms: it offers self-synchronization and self-identification capabilities, and is able to operate at ultra-low Signal-to-Noise Ratios (SNR). The ambition for the industrial partners of this project is to provide an input to the 3GPP standardization committee built from the results of the QCSP project. These results are expected, at least, to initiate discussions around the efficiency of such a solution for IoT and to push CCSK-NB-Code in a future 3GPP release, if the project is able to rally enough support from other industrial partners. To fulfill this goal, the project has 4 main objectives: Objective 1: Construction of good Non-Binary-CCSK frame for low to ultra-low coding rate (rate 1/3 down to rate 1/256). Three families of Non-Binary error control code will be studied and compared: NB-Turbo code, NB-LDPC code and NB-Polar Code. Objective 2: Define efficient detection and synchronization algorithm for a CCSK-NB-Code frame considering the whole frame itself as a preamble thanks to its particular structure. Objective 3: Development of two demonstrators, the first one using GNU radio module to perform real time demonstration, and the second one integrating new modules in Open Air Interface, the open source software, capable of emulating a complete 4G network, from the core network to the user equipment. Objective 4: The last objective is the dissemination of the results of the project with a particular focus on the 3GPP standard.