Master's Thesis: AI-Native Super-UE for 6G

Future wireless networks could allow phones, smart glasses, watches and other wearables to act as a single coordinated terminal, sharing communication, computing and sensing resources. Such cooperation can improve connectivity and distributed AI inference, but also introduces energy, signaling and synchronization costs. This thesis will develop an AI-native framework for deciding when devices should cooperate, which partners to select and how much information and computation to share.

The assignment

• Develop compact semantic representations that can be progressively refined to exchange task-relevant information at levels of detail suited to different device capabilities.
• Design an uncertainty-aware, learning-based controller that selects cooperation partners and determines the extent of information and computation exchange.
• Account for probing, communication, processing and coordination costs through a semantic token tax when assessing the net value of cooperation.
• Use modeling and system-level simulations to evaluate inference accuracy, user experience, communication efficiency, latency and energy consumption, and determine when cooperation is beneficial.
• Deliver a reproducible framework and a proof-of-concept demonstration.

Who we are looking for

• Master’s student in computer science, electrical engineering, mathematics or a related discipline.
• Strong knowledge of AI and machine learning, with an interest in signal processing and wireless communication.
• Ability to program in Python and experience using PyTorch or TensorFlow.

Practical details

Location: Lund. Preferred start: as soon as possible in 2027.