Understanding the radio frequency (RF) channel is essential for wireless communications. In complex, non-line-of-sight environments, such as dense urban, the RF propagation channel is dominated by multiple reflections, scattering, and diffraction. Hence, computationally inexpensive theoretical and empirical models cannot accurately predict the radio channel in such a scenario; the industry standard for such complex environments is the physics-based ray-tracing technique. However, ray-tracing is a computationally costly procedure even with the latest acceleration techniques, which limits its utility in real-time applications. A capability to rapidly compute channel estimates would be an enabler for a number of civilian and electronic warfare technologies. Machine learning methods offer an alternative route to formulating propagation models that can potentially combine the accuracy of physics-based models with the efficiency of theoretical and empirical models. These methods are based on training a network. While the training process is time-consuming and completed offline, once the network is trained, predictions can in principle be done in real-time.
The PhD student will implement a full 3D hybrid ray-tracing method and accelerate the ray-tracing simulations with machine learning. Other activities supporting the study will also be carried out when required. The PhD would suit an applicant with a good first degree in Computer Science, Electrical Engineering or Physics, having a good knowledge of coding and maths. We expect the PhD candidate to develop the expertise required to lead a computational research project, to train students, to interact with colleagues with different backgrounds (physics and engineering) and from different disciplines (i.e., machine learning, electromagnetism, radio-propagation).