Position location, which in general comprises Direction of Arrival (DOA) and Time of Arrival (TOA) estimation, is an important area of research in many applications, including Radar signal processing and Fifth Generation (5G) and Sixth Generation (6G) commercial wireless systems. Conventionally this has been performed using either relatively simple algorithms based on received signal power measurement or more sophisticated subspace techniques, such as MUltiple SIgnal Classification (MUSIC) and Estimation of Signal Parameters via Rotational Invariance techniques (ESPRIT). Recently, the implementation of position location estimation using Machine Learning (ML) techniques has been proposed as a methodology to improve performance beyond that of existing techniques, particularly in non-line-of-sight signal propagation conditions which occur commonly in many applications. This is an active area of research and is showing significant promise.

This project will run as part of the research council funded Centre for Doctoral Training in Sensing, Processing, and AI for Defence and Security (SPADS). This is a four year programme involving both a PhD research project and integrated studies as part of a cohort of like-minded students. The integrated studies will include advanced courses and bespoke training events such as summer schools, specialised theme meetings, and innovation and commercialisation sandpits.

The project is sponsored by the company AMD and will involve close industrial collaboration to develop and test novel machine learning algorithms for position location. Following an initial literature review to define the state of the art, the following main research phases are expected:

• Determination of the specific application of position location estimation to be addressed within the project. This will entail basic application research and determination of current state-of-the-art techniques and future directions.

• Theoretical research into conventional position location techniques based on power measurement and subspace techniques. Here the objective is to establish performance and computational complexity benchmarks for a range of conventional techniques.

• Development of machine learning models suited to position location estimation and software implementation of these based on AMD computer processor and graphics processing unit (GPU) technology. These implementations will develop and optimize the machine learning models to determine relative performance and computational complexity when compared to conventional methods.

It is likely that the specific objectives of the project will evolve throughout the PhD. However, AMD believes that it is important that the project comprises both theoretical study of the candidate techniques and practical implementation on AMD hardware. Interested candidates are requested to apply as soon as possible.