Deep Learning-Driven Traffic and Trace Generation for Beyond 5G Network Architectures and Services

• Grant number: 24/22594-4
• Support Opportunities: Scholarships abroad - Research Internship - Doctorate
• Start date: April 01, 2025
• End date: March 31, 2026
• Field of knowledge: Engineering - Electrical Engineering - Telecommunications
• Principal Investigator: Christian Rodolfo Esteve Rothenberg
• Grantee: Ariel Góes de Castro
• Supervisor: Marco Chiesa
• Host Institution: Faculdade de Engenharia Elétrica e de Computação (FEEC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil
• Institution abroad: KTH Royal Institute of Technology, Sweden
• Associated to the scholarship: 24/05363-9 - Innovative approaches for network analysis and optimization: leveraging deep learning and programmable hardware, BP.DR

Abstract

The analysis of network data is critical for understanding network behavior, identifying anomalies (e.g., traffic spikes, packet loss), and assessing user experience on network services. These services are directly impacted by the underlying network infrastructure, which must manage user requests under constraints such as latency, congestion, and dynamic traffic demands. The availability of diverse network traces enables the development and validation of solutions to anticipate these challenges across various scenarios, improving service quality and performance.Despite this need, there is a shortage of public datasets that reflect real-world conditions, particularly in academia. This shortage is especially acute for emerging applications in 5G/6G networks, such as virtual reality (VR), augmented reality (AR), and unmanned aerial vehicles (UAVs), which each bring unique requirements. The lack of representative data hinders our understanding of these applications' performance under realistic network conditions, complicating efforts to design and test effective solutions (e.g., protocols and routing algorithms).To address these gaps, this FAPESP BEPE proposal aims to evaluate neural network architectures for data generation. These architectures could serve as the foundation for tools that allow, for instance, users to selectively generate data distributions (e.g., video streaming, cloud gaming) for testing. This flexibility would support the development of resilient protocols optimized for the heterogeneous demands of future applications, such as low latency and high bandwidth.By advancing synthetic data generation methodologies, we aim to overcome data scarcity issues and foster innovation in network simulation, paving the way for robust solutions to meet the complex demands of next-generation applications.