Overview
This project is aligned with the prestigious EPSRC Centre for Doctoral Training in Distributed Algorithms, offering advanced, cohort-based training in next-generation data science.
About this opportunity
Based at the University of Liverpool, the successful candidate will gain hands-on experience with cutting-edge supercomputing facilities and work within the Signal Processing Group, collaborating with experts across areas such as Bayesian methods, machine learning, image and radar processing, data fusion, and energy-efficient computing.
Key highlights include:
- High-impact research on positioning in GNSS-denied and adversarial environments
- Advanced techniques using particle filters, machine learning, and distributed data fusion
- Dstl collaboration with co-supervision, access to real-world data, and a 3–6 month placement
- Aligned to the EPSRC CDT in Distributed Algorithms, offering cohort-based, interdisciplinary training
- Expert academic supervision from leaders in information fusion and positioning
- Access to cutting-edge GPU and supercomputing facilities at the University of Liverpool
- Scalable research scope: From single-platform to swarm-level multi-platform systems
It’s increasingly often the case that systems need to know where they are when GNSS (e.g. GPS) is unavailable. In such settings, a combination of sensors (e.g. accelerometers, gyroscopes, terrain map matching, intercepted signals from non-GNSS satellites and data from surveillance sensors configured to perform Simultaneous Localisation and Mapping, SLAM) can be used to enable a single platform to understand how its position and orientation changes over time.
While solving the positioning problem in the context of a single platform is a subject of current research and development, the focus here is on the longer-term research challenge of tackling the problem in situations involving multiple platforms. Indeed, it is often the case that, to achieve their (sometimes shared) objective, multiple platforms need to understand their position relative to each other’s current position and/or to capitalise on one another’s sensors to accurately estimate their own position relative to an absolute reference frame (e.g. such that they can guarantee relative positioning accuracy relative to another reference frame that is external to the co-operating platforms).
This project aims to develop techniques that can capitalise on data from multiple platforms to enable those platforms to provide accurate positioning relative to one another and relative to an absolute reference frame. This process is made particularly challenging in situations where GNSS is unavailable when an adversary deliberately attempts to disrupt the use of GNSS. In those settings, it is important to operate in a distributed setting whereby each platform can operate autonomously but also has the ability to capitalise on succinct summaries of other platforms’ data when they are available.
The focus of the PhD will then be on understanding what those summaries should be and how they should be processed. This is challenging because of the interplay between two competing considerations. First, there is a need to get as close as possible to the performance of a theoretically optimal centralised algorithm that would
summaries received from two platforms that have exchanged summaries between themselves already.
The proposed solution is to develop an optimal centralised solution and then use machine learning to learn the definition of the summaries (and thereby how they should be processed) with an objective function that explicitly aims to ensure that these two challenges are met simultaneously. The PhD will begin with a focus on single-platform multi-sensor setting, using state-of-the-art particle filter algorithms. The PhD will then progress to consider a centralised multi-platform setting before focusing on the motivating challenge of multi-platform distributed positioning, initially in the context of a handful of platforms and aspirational with a focus on a swarm of perhaps hundreds of platforms.
The project is affiliated to an ongoing project with Dstl such that Dstl will co-supervise the project alongside Prof Simon Maskell and Prof Jason Ralph, who are respectively experts in information fusion and positioning. It is anticipated that significant interaction with Dstl will take place and that Dstl will help provide real data to support the testing of algorithms developed in the PhD. In addition, it is anticipated that the student will spend 3-6 months gaining valuable experience of working at Dstl as part of the PhD.
The successful student will be based at the University of Liverpool and be aligned to the CDT and Signal Processing Group .
The project is open to UK nationals only.
Who is this opportunity for?
The ideal candidate will hold an undergraduate or master’s degree in a numerate subject, with a keen interest in next-generation data science, computing, and collaborating with industry partners to solve real-world challenges.