The research activities carried out by the TELICE group focus mainly on the reliability and connectivity of communications in transport systems in relation to societal and industrial needs.
These activities are part of the IEMN's flagship projects: TRANSPORT, IoT MAKE SENSE and TELECOM UHD.
Five main themes are studied:
- Contribution to the optimisation of the PHY 5G layer for vehicular communications :
To boost 5G performance in terms of latency, spectral efficiency and the number of users, 3GPP Release 18 proposed the following measures
in October 2022 to explore the potential of AI/Machine Learning to optimise channel estimation, resource allocation [...IEEESys23] and positioning, the common factor in these 3 objectives being the propagation channel. Since 2020, the TELICE group has been anticipating this research [Sensors24]. The originality of the work lies in i) the multidimensional characterisation (time-space-frequency-polarisation) of the channel in suburban and tunnel environments thanks to the numerous measurement campaigns carried out with the MAMIMOSA channel sounder, ii) the analysis of the temporal and spatial variability of the MIMO channel, iii) the creation of a database of channel parameters and iv) the adaptation of machine learning models for the detection of types of propagation scenario (LOS/NLOS). It should be noted that the recent state of the art on this subject only reports results in an indoor environment.
- IoT robustness to EM jammers and interference :
IoT systems are increasingly being integrated into critical infrastructures such as energy networks.
The SNCF, for example, is adopting LoRa technology to optimise train maintenance. Sensors on board trains monitor the condition of equipment, transmitting the data to a railway IoT platform via LoRa-LTE gateways installed on each carriage. This allows remote access to sensor data to plan any necessary interventions before arriving at the station or maintenance centre. However, IoT systems are vulnerable to cyber attacks, including LoRa attacks such as sniffing and jamming. Reactive jamming attacks, synchronised with target transmitters, are particularly effective in disrupting LoRa communications. We have studied the detection, classification and impact of these attacks on the performance of LoRa systems, based on the development of a test platform. In addition, as part of a project with the SNCF, we were able to issue recommendations for a harmonised test methodology for comparing railway wireless communication technologies under transient interference [IEEETransEMC23].
- Reconstruction of EM fields for public exposure using AI
Exposure to electromagnetic fields (EMF) is receiving a great deal of attention in light of the 5G network. Despite this, accurate reconstruction of the electromagnetic field across a region remains difficult due to a lack of sufficient data. TELICE's original research focused on exploring reconstruction strategies based on observations from certain localised sites or spatially distributed sensors, using techniques based on geostatistics and Gaussian processes. Recent initiatives have focused on the use of Conditional Generative Adversarial Network (cGAN) machine learning conditioned by the topology of the environment and artificial intelligence for this purpose [...].Sensors22]. A promising approach allows the reconstruction of EMF exposure maps without using a large training set by deriving a neural tangent kernel for an infinitely large neural network from a sparse set of values measured by sensors.
- Design and implementation of multiple Terahertz transmitters using a combination of multi-core fibre and photodiode arrays: application to very high-speed telecommunications
With the aim of enabling very high-speed communications with improved link budgets, the work focused on bringing together two different technological building blocks: 1) an array of THz transmitters made up of photodiodes which, thanks to the principle of power combination well known in radio frequency, enables the system to generate more THz power than conventional unitary transmitters, and 2) multi-core fibres to excite the photodiode array, which are also of interest for achieving spatial multiplexing in future optical networks. For the first time, the excitation of an array of 4 photodiodes by a 4-core fibre has been demonstrated, opening up the possibility of making links around 300 GHz at data rates of up to 100 Gbit/s, with better link budgets than those of links based on a single photodiode [OFC22]. Alternatively, it would be possible to perform spatial data multiplexing on 4 different THz carriers in order to achieve data rates of several hundred Gbit/s.
- Diagnosis of on-board wired networks
The challenge is to develop an original system for monitoring the state of health of wired communication and energy networks in vehicles by combining two approaches, a signal approach and a system approach. Thanks to collaboration with the CRISTAL laboratory, an original method for detecting and isolating non-hard faults has been proposed. In contrast to reflectometry, which is widely used in fault detection systems, this transferometry method is based on the estimation of network transmission coefficients, and the calculation of residuals that are more sensitive to faults, and on the decomposition into Y-shaped sub-networks [ISATrans23]. The method has been validated on a laboratory test bench.
