PhD Studentship: Distributed and Lightweight Large Language Models for Aerial 6G Spectrum Management

About the Project

Project details:

Aerial platforms such as Unmanned Aerial Vehicles (UAVs), High-Altitude Platforms (HAPs), and Low Earth Orbit (LEO) satellites will play a vital role in the next generation of wireless communication (6G) to extend network coverage, supporting diverse data-intensive applications such as immersive extended reality and autonomous systems. However, aerial 6G networks will operate in spectrum environments that are scarce, heterogeneous, and highly dynamic, which makes the traditional static and centralised spectrum management strategies inadequate for ensuring reliable, low-latency, and scalable operation in aerial 6G networks. In this regard, Large Language Models (LLMs) have recently emerged as a key technology to achieve adaptive 6G spectrum management. 

The core idea of LLM is to exploit transformer-based attention mechanisms to model sequential dependencies and capture long-range interactions, making them promising tools for complex spectrum management. Despite being such a promising technology, the centralised and resource-intensive nature of current LLMs conflicts with the constraints of aerial 6G networks in terms of limited computation, energy, and communication resources. To fill this gap, this proposal aims to design novel distributed and lightweight LLMs for spectrum management in aerial 6G networks. Specifically, the project will design wireless-aware data representations and embedding techniques that allow LLMs to natively interpret spectrum-related information, capturing the unique temporal, spatial, and physical characteristics of 6G signals. Building on this foundation, this project will advance distributed and lightweight LLM architectures to enable resource-efficient spectrum allocation across highly dynamic and interference-prone 6G aerial environments. As LLM-enabled 6G is a nascent field, this project is at a pivotal time to advance our knowledge on how to build distributed and lightweight LLM architectures for 6G spectrum management. The success of this project will contribute to achieving the UK’s "Connected Nation" ambition by advancing the integration of LLM in 6G systems. 

Three cooperative Work Packages (WPs) are envisioned to achieve the research aim. WP1: Wireless-aware Data Representations and Embedding (Months 1–12): This WP will first design a structured data representation approach to encode spectrum information into token sequences compatible with transformer-based architectures. Then, a novel embedding technique will be designed to map high-dimensional wireless network parameters into a compact vector space modelling, where temporal, spatial, and spectral correlations will be effectively captured in model training. WP2: Distributed LLM-enabled Spectrum Optimisation (Months 13–26): Building on the pretraining foundations from WP1, this WP will first exploit parameter-efficient fine-tuning techniques, e.g., pruning and quantisation, to reduce the complexity of LLMs. Then, distributed inference strategies and consensus mechanisms will be investigated to coordinate spectrum allocation decisions across distributed 6G devices, ensuring consistency, scalability, and conflict-free operation in highly dynamic and interference-rich environments. WP3: Algorithm Validation and Use Case Demonstration (Months 27–36): This WP will first develop an integrated hardware–software testbed to systematically validate the performance of proposed solutions under diverse channel conditions, mobility patterns, and interference dynamics. Then, two representative use cases, including emergency response networks and rural connectivity provision, will be implemented to showcase the project’s innovations for future 6G networks. 

Please direct project specific enquiries to: Dr Wang Miao (wang.miao@exeter.ac.uk) Please ensure you read the entry requirements for the potential programme you are applying for. To Apply for this project please click on the following link -https://www.exeter.ac.uk/study/funding/award/?id=5733

Advert information