Thèse "LLMs : assistants d'optimisation de l'allocation de ressources dans les réseaux futurs" F/H

Descrizione del lavoro:

Publication date : Apr 10, 2026, 5:52PM
As mobile networks evolve toward 6G, managing Radio Access Network (RAN) resources is becoming increasingly complex. This stems from the need to serve massive heterogeneous users with diverse QoS requirements, stringent latency constraints, and strict energy targets, all while operating under physical, spectral, and computational limitations. This has made resource allocation optimization in RAN a critical and computationally intensive challenge. The optimization problems in RAN are often non-convex due to inter-cell interference, nonlinear QoS models, scheduling strategies, and power constraints in coupled systems. Traditional methods struggle with scalability, slow convergence, and poor adaptability to dynamic environments. Reinforcement learning offers adaptability but demands extensive training and is sensitive to changing states, limiting its reliability in real deployments.
LLMs have shown remarkable problem-solving capabilities beyond text and are beginning to assist traditional symbolic solvers. The objective of the thesis is to investigate LLMs as assistants in the optimization of RAN. The study explores integrating LLMs within traditional optimization, leveraging their ability to interpret and transform complex problems into more tractable forms while providing explainable, real-time solutions.
The integration presents a transformative approach to address complex, non-convex optimization problems, introducing technical and practical challenges that must be navigated to fully realize their potential.
The proposed framework capitalizes on LLMs' reasoning abilities to transform application-level objectives into optimization problems and to apply transformation techniques such as convex relaxation, dual decomposition, etc. LLMs will be trained to identify and apply suitable transformations based on the problem's characteristics and the required solution accuracy, and then to rely on traditional solvers selected from the broad range of optimization methods developed by the networking community. To ensure the feasibility and robustness of the solutions, iterative verification and refinement processes will be integrated, incorporating feedback mechanisms, uncertainty quantification, and probabilistic verification to dynamically adapt and improve the proposed solutions.
Some prior works explore LLM-based methods for wireless resource optimization but do not fully address critical real-time constraints, latency, or realistic convergence challenges, whereas this thesis proposes solutions tested across diverse 6G scenarios with comprehensive evaluation in both simulated and realistic environments.
In conclusion, while integrating LLMs into RAN resource allocation presents a promising direction for tackling complex, dynamic optimization problems, overcoming challenges related to feasibility, real-time operation, and convergence is essential. This thesis will unlock the potential of LLMs as assistants for future wireless network optimization.
o Master in telecommunications or applied mathematics
o Learning skills(reinforcement learning, Markov decision processes, ...)
o Skills in optimization
o Skills in modeling networks and systems(probabilities, random processes, queues, ...)
o In-depth knowledge in the field of fixed and mobile networks (architectures and protocols, virtualization, SDN, 5G, quality of service management, resource allocation, etc.)
o Programming and IT development skills (Python, Matlab, C / C ++, ...)
o Fluency in written and oral English
o An internship experience in the field of artificial intelligence would be a plus.
The doctoral student will join Orange, an international group, expert in the information and communication technologies sector. In particular, he will join the MORE team made up of application and research engineers in applied mathematics and AI within the Data-AI department. This team is in charge of developing models and techniques to optimize the use of resources (in the broad sense) and to assess network performance. He/she will work in an international context and will have the opportunity to contribute to research projects with external partners from the academic and industrial environment.

In addition of the gross salary, the company offers a corporate savings and retirement plan, profit-sharing, employee stock ownership, health and life insurance coverage, discounts on Orange products, as well as social and cultural activities.
Orange Innovation brings together the research and innovation activities and expertise of the Group's entities and countries. We work every day to ensure that Orange is recognized as an innovative operator by its customers and we create value for the Group and the Brand in each of our projects. With 720 researchers, thousands of marketers, developers, designers and data analysts, it is the expertise of our 6,000 employees that fuels this ambition every day.
Orange Innovation anticipates technological breakthroughs and supports the Group's countries and entities in making the best technological choices to meet the needs of our consumer and business customers.

Within Innovation, you will be integrated into a research team at the forefront of innovation and expertise on the networks of the future.
MORE (Mathematical Models for Optimization and Performance Evaluation) team is responsible for research in applied mathematics and AI, to develop models and techniques for resource optimization and infrastructure performance evaluation network. Research activities are carried out to identify technological weaknesses and design analytical tools for the design and planning of these infrastructures.
At Orange, only your skills matter.
Regardless of your age, gender, background, origin, religion, sexual orientation, disability, neurodiversity, or appearance, we actively encourage diversity within our teams, as it is a collective strength and a driver of innovation.
Orange is a disability-inclusive employer: please feel free to let us know about any specific needs you may have