Master Thesis: Quality of Service for AI Token Traffic in 6G Networks

Join our Team

About this opportunity:

The rapid growth of AI services is creating new types of network traffic. AI inference generates output as streams of discrete tokens, which differ significantly from traditional video or web traffic in terms of traffic patterns, latency requirements, and sensitivity to packet loss. Understanding and supporting this new traffic class is becoming increasingly important as AI workloads move to the network edge and cloud. 3GPP has recognized this trend and initiated studies on how mobile networks can better support AI-related traffic. Several open questions remain around how to characterize token traffic, what QoS requirements different types of AI services have, and whether existing 5G QoS mechanisms are sufficient or need to be extended.

What you will do:

The goal of this thesis is to investigate the QoS requirements of AI token traffic in mobile networks and evaluate how well existing and potential new mechanisms can support this traffic class. The following steps are envisioned as part of the thesis work:

1. Literature study: Review the current state of the art in AI traffic characterization, 3GPP QoS mechanisms (5QI, PDU Sets, GBR/non-GBR bearers), and related academic work on supporting emerging traffic types in 5G/6G networks.
2. AI traffic characterization: Study the properties of token traffic generated by different AI services (e.g., LLM chat, AI agents, neural codecs). Characterize traffic patterns such as token rates, burst behavior, and latency sensitivity using open-source AI models.
3. Simulation and evaluation: Develop or extend simulations to model AI token traffic over a mobile network. Evaluate how existing QoS mechanisms perform for this traffic class and identify potential gaps.

The skills you bring:

This project is aimed at students in electrical engineering, computer science, computer engineering, or similar. The following background is preferred:

• Understanding of mobile network architecture (4G/5G) and QoS concepts
• Familiarity with machine learning concepts, particularly large language models
• Programming experience in Python; experience with C++ is a plus
• Interest in 3GPP standardization and telecom research

Extent: 1 student, 30hp

Location: Stockholm, Sweden

Preferred Starting Date: Nov 1, 2026

Keywords: 5G/6G, AI Traffic, Quality of Service, Large Language Models, Token Communication, Network Simulation, 3GPP