This project will explore how learning and reasoning should be applied to radio resource and topology management strategies in 5G and beyond 5G heterogeneous networks, while ensuring that they remain robust to mobile device malfunction and system attacks, which exploit the distributed nature of the control. Strategies could include machine learning and other artificial intelligence approaches, possibly applied using game theoretic techniques. The purpose of the project will be to show how, by applying these forms of artificial intelligence, it is possible to improve the robustness, flexibility and usage of pooled radio spectrum, while generating more energy efficient network topologies, both on a local and system wide basis. The project will establish where the learning/reasoning should best reside (nodes and/or network), and also the degree of control information exchange required between nodes. A mixture of simulation and analysis will be used to assess performance.