My research investigates how federated learning (FL) can be efficiently implemented over resource-constrained, multi-layer wireless networks with heterogeneous computing and communication capabilities. I am particularly interested in understanding the interactions between learning performance, communication constraints, computational resources, and system-level costs.

A more recent direction of my work explores the programmability and softwarization of federated learning, with the goal of enabling Federated Learning as a Service (FLaaS). This involves investigating whether different stages of the FL process can be represented as modular network functions and dynamically deployed across multi-layer edge–cloud infrastructures.

My longer-term objective is to develop frameworks for FL function placement, service chaining, and resource allocation that account not only for conventional network costs, such as latency and energy consumption, but also for learning convergence and model performance.

My research interests include radio access network architectures, network function placement, and the integration of machine learning into wireless network operations.

Building on my earlier work on RAN function placement and my involvement in the 5G-STARDUST project, I am interested in exploring how distributed learning and optimization techniques can be implemented within Open RAN (O-RAN) architectures.

My objective is to investigate learning-based approaches for network resource management, service orchestration, and RAN automation, considering heterogeneous network conditions and end-user service requirements.

My research focuses on multi-objective and multi-time-scale optimization for intelligent resource management across heterogeneous, multi-layer edge–cloud computing environments.

A key challenge is the joint optimization of service placement, computing-node selection, and computational offloading, where decisions operate at different time scales and influence one another. My previous work has addressed this challenge through multi-time-scale Markov Decision Process (MDP) formulations and reinforcement learning, particularly in vehicular and non-terrestrial networking scenarios.

Building on these contributions, I aim to develop scalable optimization and learning-based frameworks that jointly consider latency, energy consumption, resource utilization, and service performance under dynamic network conditions and heterogeneous computing capabilities.

6G and Beyond-5G Wireless Networks

Distributed intelligence, AI-driven network optimization, and intelligent resource management for next-generation wireless networks, with an emphasis on programmable architectures, heterogeneous network infrastructures, and efficient integration of communication, computing, and learning.

Non-Terrestrial Networks and Satellite Computing

Resource management, distributed learning, and computational offloading across satellite–terrestrial and space–edge–cloud infrastructures, with particular interest in low-Earth-orbit satellite networks.

Connected Vehicles and Intelligent Transportation Systems

Distributed intelligence, collaborative learning, and resource optimization for connected vehicles and Internet of Vehicles applications, including edge-assisted computation and intelligent service management.