Module Directory

This module explores how state-of-the-art AI techniques, such as deep learning, reinforcement learning and federated learning, can be applied to the design and operation of next-generation telecommunication networks. Building on prior knowledge of machine learning and modern communication principles, it shows how key network problems (primarily at the MAC and PHY layers) can be cast as learning and decision-making tasks, and how AI-based solutions can improve performance and automate tasks. Through practical work with varying network models, students develop both the theoretical insight and hands-on skills needed to engage with AI-native communication networks and systems in research and industry. The module is designed for students who already have a basic understanding of communication networks and wish to deepen their knowledge by examining AI-driven approaches.

The aims of this module are:

• To develop a clear and rigorous understanding of how modern AI techniques, particularly deep learning, reinforcement learning and federated learning, can be applied to problems in next-generation telecommunication networks.


• To enable students to formulate key PHY- and MAC-layer tasks as learning and decision-making problems, and to recognise when a particular AI approach is appropriate.


• To provide hands-on experience in designing, training and evaluating AI-based models and controllers using suitable software tools and network models.


• To cultivate a critical awareness of the opportunities and limitations of AI-native communication systems, including considerations of data availability, scalability, reliability and deployment.

By the end of this module, students will be expected to be able to:

1. Explain and compare key AI paradigms relevant to next-generation telecommunication networks


2. Formulate telecommunication problems as learning or decision-making tasks and design, implement and train learning models for these tasks


3. Analyse learning–based models for control or optimisation problems in telecommunication networks and evaluate their performance using appropriate metrics.


4. Configure and implement a learning approach in telecom-relevant setting, and analyse trade-offs among learning performance, communication cost and practical constraints.


5. Critically evaluate AI-based solutions for next-generation telecommunication networks, considering issues such as robustness, scalability, data requirements and deployment constraints.

Module information (Outline Syllabus)

• Overview of AI in next-generation telecommunication systems: motivation, scope, and examples of AI-driven PHY/MAC functionalities.


• Foundations of learning-based communication systems: problem formulation, supervised vs. reinforcement vs. federated learning in telecom contexts.


• Deep learning for wireless communications: neural architectures for tasks such as estimation, detection, prediction and adaptation.


• Model development and evaluation: training workflows, dataset considerations, performance metrics, and comparison with classical signal-processing-based methods.


• Reinforcement learning fundamentals: Markov decision processes, value and policy-based methods, and practical considerations for network control tasks.


• RL applications in simplified telecom scenarios: power control, resource scheduling, mobility/handover decisions, energy optimisation and related control problems.


• Federated learning principles: decentralised learning, aggregation methods, client heterogeneity, and communication–computation trade-offs.


• Federated and distributed learning for telecom settings: edge constraints, device–network interaction, privacy considerations and practical deployment issues.


• AI-native communication architectures: opportunities and limitations of AI-driven PHY/MAC designs, including robustness, scalability and data requirements.


• Practical development using software tools: implementation of learning-based models and controllers using Python (or any other preferred language/tool), ML libraries, and simplified wireless/network models.

This module will be delivered via:

• One 2-hour lecture per week
• One 2-hour lab / in-class test) per week