Artificial Intelligence research at L2S

The L2S Artificial Intelligence initiative coordinates AI research activities across L2S’s three research areas: Systems and Control, Telecommunications and Networks, and Signal Processing and Statistics.

The laboratory’s approach builds on its long-standing strengths in control theory, information theory, signal processing, statistics, optimization, and uncertainty quantification. These disciplines provide a natural methodological foundation for developing reliable and explainable AI methods suited to physical systems, networks, signals, and complex data.

The initiative aims to coordinate research activities, projects, recruitment, and education related to AI, in connection with CentraleSupélec, Université Paris-Saclay, Institut DATAIA, and national research networks in the field.

Scientific scope

Activities are structured around five complementary directions, ranging from the mathematical foundations of AI to applications in systems, communications, energy, health, and industry.

AI for dynamical systems and control

Learning for dynamical systems, optimal control, system identification, safety of cyber-physical systems, and data-driven control.

AI for signals and images

Statistical learning for signal processing, inverse problems, imaging, computer vision, time series, and multimodal data fusion.

AI for communications, energy, and networks

Communication networks, 5G/6G, resource allocation, smart grids, sensor networks, and energy systems.

Robust AI and uncertainty quantification

Probabilistic methods, surrogate models, calibration, conformal prediction, and uncertainty quantification for assessing the reliability of AI models.

Sequential learning and decision-making under uncertainty

Multi-armed bandits, reinforcement learning, structured Markov decision processes, Bayesian optimization, and the design of computer experiments.

AI across all three research areas

Systems and Control

Learning for dynamical systems, optimal control, robotics, safety of cyber-physical systems, digital twins, and integration of physical models and data.

Telecommunications and Networks

AI for communications, resource allocation, complex networks, distributed learning, sequential decision-making, and optimization for digital infrastructures.

Signal Processing and Statistics

Statistical learning, signal processing, inverse problems, Gaussian processes, probabilistic models, uncertainty quantification, and robust AI.

Organization and scientific activities

The L2S Artificial Intelligence initiative is embedded in a rich scientific environment within CentraleSupélec and Université Paris-Saclay. L2S members contribute in particular to Institut DATAIA, the ISN Graduate School, GIS UQ@Paris-Saclay, the H-Code interdisciplinary initiative on human-machine hybrid systems, and national networks such as GdR IASIS, GdR MACS, and RT CNRS UQ.

The laboratory also contributes to scientific activities in the field through regular seminars, thematic workshops, and national and international collaborations. The L2S AI Day held in 2024 helped formalize the scientific scope of the initiative and strengthen exchanges among the three research areas.

Recent scientific activity

Over the 2023-2025 period, AI accounts for a substantial and growing share of the laboratory’s activities, with projects, PhD theses, and hiring across all three research areas. Analyses of 2024-2025 HAL publications show a balance between applied AI and methodological contributions in machine learning, optimization, mathematics for AI, statistics, and uncertainty quantification. The share of AI-related publications rose from approximately 33% in 2024 to 41% in 2025.

	Systems and Control	Telecommunications and Networks	Signal Processing and Statistics
AI hires, 2023-2025	1	3	3
AI PhD theses 2023-2025	14	47	24
AI projects 2023-2025	11	6	10

These figures come from the synthesis work prepared as part of the development of the L2S Artificial Intelligence initiative.

Objectives

The goal of the L2S Artificial Intelligence initiative is to consolidate L2S research in artificial intelligence by building on the laboratory’s long-standing strengths: applied mathematics, systems science, information processing, optimization, and probabilistic modeling.

strengthen L2S’s visibility in AI research at the interface between methodological foundations and applications;
foster collaborations among the three research areas on shared research questions;
support AI projects, recruitment, and education initiatives;
develop reliable and robust AI methods suited to the complex systems studied at the laboratory.