Data Driven Approaches for Machine Learning in Wireless Communications

Fifth Generation (5G) wireless networks are starting to be deployed around the world, in order to provide higher capacity and more energy efficient communications in the future. These standards exploit large scale antenna arrays to perform very directional beamforming between base stations and mobile users. At the same time, machine learning methods are enabling a revolution in data processing that can provide solutions to many problems that were previously thought to be too complex to solve. These approaches can identify complex interactions between different parts of the available data during a training phase, which can then be exploited in practical operation.
Among wireless communications researchers, there is a strong desire to apply machine learning to improve the performance of future networks. Firstly, it is very important to identify problems where these methods can give system designers an advantage over traditional signal processing algorithms. There is a second major challenge, which is to develop new methods that can achieve enhanced performance but without excessive implementation complexity. This requirement is particularly true for mobile handsets, whose performance is limited by the battery that powers it.  
The goal of this research project is to study how machine learning can improve the physical transmission and reception of data in a wide range of radio environments. Current wireless standards, such as 5G communications, embed known waveforms in the data to enable the receiver to determine the radio channel. This task enables the receiver to recover transmitted data efficiently, but it becomes challenging in the presence of co-channel interference from nearby transmitters operating on the same frequency. This project will investigate machine learning approaches to this problem and will focus on how such networks can learn from realistic datasets of communication signals. The PhD research will focus initially on three main aspects: dataset construction; evaluation of methods to identify the current scenario under which the receiver is operating and thirdly a detailed performance evaluation of these approaches.  

Closing Date: 

Wednesday, September 30, 2020

Principal Supervisor: 

Assistant Supervisor: 


An undergraduate degree in Electronic and Electrical Engineering, Computer Science or related subjects.

Minimum entry qualification - an Honours degree at 2:1 or above (or International equivalent) in a relevant science or engineering discipline, possibly supported by an MSc Degree. Further information on English language requirements for EU/Overseas applicants.


Applications are welcomed from self-funded students, or students who are applying for     scholarships from the University of Edinburgh or elsewhere.

Other funding arrangements: We are currently arranging 50% funding of the PhD project with an industrial sponsor. If the project proceeds well, the company will extend the funding to a full PhD studentship.

Further information and other funding options.

Informal Enquiries: