Splitfed with Contribution-Aware Client Selection for Mobile Telecom Environments

• Grant number: 26/16721-9
• Support Opportunities: Scholarships abroad - Research Internship - Master's degree
• Start date: October 01, 2026
• End date: March 31, 2027
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computer Systems
• Principal Investigator: Luiz Fernando Bittencourt
• Grantee: Camilo Henrique Martins dos Santos
• Supervisor: Pedro Porto Buarque de Gusmao
• Host Institution: Instituto de Computação (IC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil
• Institution abroad: University of Surrey, England
• Associated to the scholarship: 25/02185-5 - Federated Learning Architectures for Telecom Operators, BP.MS

Abstract

The growing adoption of mobile edge devices, vehicles, drones, and IoT systems in 5G networks creates new opportunities for distributed artificial intelligence. However, federated learning in these environments faces challenges related to mobility, resource heterogeneity, communication constraints, and client reliability. This proposal investigates adaptive client selection mechanisms for Split Federated Learning (SplitFed) in multi-operator 5G networks. The proposed framework combines the privacy benefits of Split Learning with the scalability of Federated Learning while dynamically selecting clients based on telemetry and Quality of Service (QoS) indicators, including mobility, resource availability, network conditions, and reliability. By filtering and selecting clients according to current operating conditions, the framework aims to reduce communication overhead, mitigate straggler effects, improve convergence speed, and preserve QoS during mobility events. The research will be integrated into the MAESTRO Federated Learning as a Service (FLaaS) platform developed within the SMARTNESS ERC and validated under realistic distributed learning and mobile networking scenarios. The expected outcomes include more efficient and robust SplitFed training, improved client participation management, and scalable client selection strategies for future 5G and 6G distributed learning environments. (AU)