Imaging, Data and Communications

Autonomous Sensing Platforms

Resilient PNT as a Service using multi-orbit satellite signals of opportunity

Research Themes

Sensor Signal Processing

Aim

Optimise PNT as a Service using satellite signals of opportunity (SoOP) across LEO/MEO/GEO.

Objectives

Establish a testbed that will enable PNT services using satellite SoOP.
Benchmark the performance of PNT services using satellite SoOP against GNSS-based systems.
Optimise data management of a the system for performance and robustness
Develop AI-driven models for the selection of the optimum SoOP to be used in a PNT as a Service scenario

Description

GNSS-enabled positioning, navigation, and timing (PNT) is under increasing threats from interference and spoofing.

Satcom systems offer signals of opportunity (SoOP) that can be used to deliver PNT services. They come from an abundant number of satellites across LEO/MEO/GEO, each offering distinct advantages in terms of signal strength or coverage. SoOPs are also available across a wide range of frequencies (from C-band to Ka-band) and multiple angular directions, thus making SoOPs a highly reliable option of pervasive PNT services.

This project will access a UK-wide network of monitoring and reference stations publishing SoOPs across C-, Ku- and Ka-bands. Data from this network will be used in conjunction with signals captured at a user terminal to deliver and benchmark PNT. The data management aspects of such a system will be investigated to optimise performance. AI-driven algorithms for selecting optimum usage of the proposed system will be developed and evaluated.

Industry Partner

This project will be in collaboration with PNTaas

https://space.site.hw.ac.uk/

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.

Full funding is available for this position.

31 Jan, 2026

Assistant Prof Kieran O'Leary

Prof George Gousettis

spads@ed.ac.uk

Further information and other funding options.

Machine Learning in Dynamical Systems for Sensor Signal Processing

Dynamical system models have been the main pillar of conventional model-based approaches in control, signal processing and sensor fusion: Sensor signal processing and inference algorithms for applications such as multi-object detection and tracking, robotic simultaneous localisation and tracking (SLAM) and calibration of autonomous networked sensors are designed by combining the known physics and stochastic elements into dynamic system models. Model inaccuracies can be mitigated to achieve significant performance gains in inference and decision-making by leveraging data and model size, following the recent advances in machine learning. However, learning from data in dynamical system models to jointly address the epistemic and aleatoric uncertainties involved remains a challenge stemming from a range of factors such as noisy data and labelling, inhomogeneous sampling, model complexity and the intractability of posterior inference.

This PhD project aims to conduct underpinning research to address problems involving the learning of hierarchical, time-varying multi-dimensional state space models for dynamic objects/phenomena, noisy measurements made in complex backgrounds and/or in the presence of calibration errors.

The incumbent will have the opportunity to steer the direction of the research in consideration of the impact on engineering problems, including learning models for complex backgrounds in radar detection, learning of birth and trajectory models to improve detection and tracking, or semi-supervised/unsupervised training of sensor data classifiers.

The application documents should provide sufficient information and evidence regarding the applicant’s background and goals, and include:

A Personal Statement
CV
Official certificates and transcripts (and English language translations, if applicable)
English Language certificate (if applicable, for more information, please see here: https://study.ed.ac.uk/programmes/postgraduate-research/947-engineering)
2 Recommendation Letters
Research proposal (optional for candidates who are not self-funded or applying for scholarships. Must be related to the project description)

Dr M Uney is an experienced scientist with interests in signal processing, machine learning, probabilistic models and Bayesian computations in sensor fusion and signal & information processing.

Prof Mike Davies holds the Jeffrey Collins Chair in Signal and Image Processing at the University of Edinburgh. His research interests are in machine imaging, data-driven computational sensing and imaging, and sensor and information fusion.

This position is open to UK/EU nationals and international applicants, and offers an annual stipend of £21,935 (as of 2024-25 fiscal year, subject to revision) for a duration of 3.5 years with £5000 research expense funds for the duration of the study.

UK nationals and other eligible applicants might choose to align this studentship with the Sensing, Processing and AI for Defence and Security Centre for Doctoral Training (SPADS CDT) at the University of Edinburgh’s School of Engineering, and benefit from an enhanced stipend (subject to CDT administration approval and security clearance, see the eligibility criteria in the link) upon acceptance. Eligible applicants should state their interest in their Personal Statement.

Minimum criteria:

- A 2:1 Master’s Degree in engineering, computer science, statistics, physics or in an appropriate subject or a related discipline.

- The University’s English language requirements

Preferable criteria:

- A 1st class Undergraduate Degree in engineering, computer science, statistics, physics or in an appropriate subject or a related discipline.

- A 1st class Master’s Degree in engineering, computer science, statistics, physics or in an appropriate subject or a related discipline.

- Experience in scholarly writing.

Tuition fees + stipend are available for Home/EU and International students

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

22 Feb, 2026

Professor Michael Davies

Dr Murat Uney

M.Uney@ed.ac.uk

Fast, explainable and modular generative AI models for computational imaging by unfolding and distillation of Langevin sampling algorithms

Research Theme

Sensor Signal Processing

Aim

To advance generative AI technology for computational imaging by developing novel neural network architectures that combine the explainability, modularity and flexibility of MCMC-based Bayesian imaging methods with the accuracy and scalability of deep learning techniques.

Objectives

Develop new neural network architectures and supervised training strategies tailored for physics-informed generative image reconstruction and uncertainty quantification, with a focus on improving accuracy, scalability to very large problems (e.g., images of size 1024x1024 pixels or larger), explainability (e.g., with layers/modules that map clearly to instrumental models, data fidelity models and regularisation models) and modularity (e.g., that allow modifying or adjusting instrumental models and noise models during inference, without need for retraining).
Leverage the proposed architectures and the industrial partner’s expertise to co-develop novel Bayesian imaging solutions for a flagship application of interest to the partner.
Develop self-supervised training strategies to fine-tune the proposed architectures directly from measurement data, bypassing the need for ground truth data.

Description

Modern computational imaging method rely increasingly on deep generative models to address challenging inverse problems. A notable example is the LATINO-PRO sampler, which combines a large-scale Stable Diffusion XL image prior trained on over five billion image–text pairs with a Bayesian inversion framework tailored to generative AI, achieving state-of-the-art performance on difficult tasks such as x16 image super-resolution. This project aims to develop novel neural network architectures specialised for generative computational imaging. A key novelty is that the proposed architectures will be derived from modern MCMC samplers such as LATINO and from recent work on unfolding, training and distilling of MCMC algorithms into physics-informed generative neural networks. The resulting models are expected to be extremely fast while preserving the structure, interpretability, and uncertainty quantification capabilities of iterative Bayesian inference. This enables flexible inference, including runtime specification of sensing models and their automatic calibration via empirical Bayesian techniques. The goal is to deliver unprecedented image reconstruction accuracy while addressing critical requirements for quantitative imaging in security and defense, notably trustworthiness and explainability. We will consider applications to thermal, multispectral, or low-photon imaging, with possible extensions to dynamic imaging and video reconstruction. For scenarios where reliable ground-truth data are unavailable for training, we will investigate pre-training strategies using public or synthetic datasets, followed by self-supervised fine-tuning directly on sensor data.

The proposed project builds on the recent paper “Learning few-step posterior samplers by unfolding and distillation of diffusion models” https://arxiv.org/abs/2507.02686, which applies deep unfolding to the LATINO sampler https://arxiv.org/abs/2503.12615, delivering remarkably accurate posterior samples in as little as 8 neural function evaluations.

A UK first-class honors degree (or its international equivalent) in computational mathematics, computer science, electronic engineering, or a closely related discipline. This project also requires strong programming skills and experience with machine learning frameworks (e.g., PyTorch).

Further information on English language requirements for EU/Overseas applicants.

Full funding is available for this position.

31 Jan, 2026

Dr Konstantinos C Zygalakis

Prof Marcelo Pereyra

Further information and other funding options.

Robust Data-Fusion for Drone Surveillance Systems Using Cross-Domain AI

Drone-based Earth observation, surveillance, and goods delivery solutions are among the recently enabled technologies that heavily rely on the reliability of on-board sensors and the integrated processing of collected data. Such sensor systems can be unreliable or occasionally unavailable in certain modalities, such as cameras and LiDAR in foggy, cloudy, or dusty environments, and radar in radio-frequency congested or denied environments.

The aim here is to conduct low-shot training of cross-domain AI models in order to: (1) improve the overall reliability of data analysis when some modalities are absent or noise-contaminated in harsh situations, and (2) enhance the overall accuracy of existing models.

We aim to extend the scientific understanding and basic technology solutions for drone-based sensor platforms operating in urban environments. These solutions have various applications in security, defence, climate change monitoring, and humanitarian domains, such as disaster surveillance, search and rescue, and urban planning surveillance.

Please note that this advert might close sooner once a suitable candidate is found. Therefore early applications are advised.

Drone surveillance has various potential applications of interest to relevant industry. While such flying objects are equipped with multiple sensors, a reliable data-adaptive sensor fusion framework is required to accommodate erroneous measurements and provide the necessary precision and reliability for the task. This PhD project specifically aimed to address the following key research questions:

“Accuracy gap”: How can data from different modalities, with varying accuracies, be combined to improve overall decision performance, such as detection, tracking, and classification accuracy?

“Reliability gap”: How can the unreliability of certain modalities be identified and mitigated using limited data from other modalities?

“Generalisation gap”: How robust is the solution to distribution shifts in the inputs, i.e., slight changes in the mode of operation such as day–night transitions?

a 1st Class undergraduate degree (or equivalent);
University’s English language requirements.

Fudning is available for home applicants only (UK+EU settled/pre-settled). For more information on this, please contact the project spervisor, Dr Mehrdad Yaghoobi (m.yaghoobi-vaighan@ed.ac.uk).

01 May, 2026

Dr Mehrdad Yaghoobi Vaighan

Professor Mike Davies

m.yaghoobi-vaighan@ed.ac.uk

Novel Computing and Beyond CMOS Hardware

High-Performance Kernel Learning Processors for 6G Sensing: From Algorithmic Models to ASICs

Research Theme

Sensor Signal Processing

Aim

High-Performance Kernel Learning Processors for 6G Sensing: From Algorithmic Models to ASICs

Objectives

Develop analytical models that characterise the stability, convergence behaviour, and error performance of kernel-based online learning algorithms under the high-speed conditions expected in 6G sensing systems.
Design and optimise high-speed kernel learning algorithms that exploit sparsity, feature selection, and reduced model complexity to enable real-time sensor signal processing.
Create hardware-efficient architectures implementing the proposed algorithms, investigating trade-offs between throughput, energy consumption, silicon area, and learning accuracy for ASIC deployment.
Validate prototype ASICs, performing real-time testing using representative 6G sensing data to demonstrate end-to-end functionality and performance.

Description

This project will also be supervised by Prof George Goussetis.

This project investigates how kernel-based online learning can operate under the extreme data rates, tight latency, and dense sensing environments expected in 6G systems. Although well suited to continuously evolving sensor data, kernel online learning is fundamentally constrained by nonlinear parameter-update loops that become computational bottlenecks at 6G-class speeds. The core problem is the absence of analytical understanding and hardware-efficient formulations that explain these limits and indicate how they can be overcome.

The research will develop models that characterize stability, complexity, and error behaviour under realistic 6G operating conditions, revealing the constraints and sparsity structures that determine real-time feasibility. These insights will guide the exploration of algorithmic variants and architectural principles capable of supporting high-speed, low-energy kernel adaptation.

Validation will use representative 6G sensing workloads, establishing a clear pathway from problem characterization to hardware-ready design principles suitable for future ASIC implementations.

[1] M. Scarpiniti et al., “Nonlinear spline adaptive filtering,” Signal Process., vol. 93, no. 4, pp. 772–783, 2013.

[2] W. Liu et al., “The kernel least-mean-square algorithm,” IEEE Trans. Signal Process., vol. 56, no. 2, pp. 543–

554, 2008.

[3] W. D. Parreira et al., “Stochastic behavior analysis of the Gaussian kernel least-mean-square algorithm,” IEEE

Trans. Signal Process., vol. 60, no. 5, pp. 2208–2222, 2012.

[4] N. J. Fraser et al., “FPGA implementations of kernel normalised least mean squares processors,” ACM Trans.

[5] M. T. Khan and O. Gustafsson, “ASIC implementation trade-offs for high-speed LMS and block LMS adaptive

[6] M. T. Khan et al., “Optimal complexity architectures for pipelined distributed arithmetic-based LMS adaptive filter,” IEEE Trans. Circuits Syst. Regul. Pap., vol. 66, no. 2, pp. 630–642, 2018.

[7] M. T. Khan and O. Gustafsson, “Stochastic analysis of LMS algorithm with delayed block coefficient adaptation,”(arXiv:2306.00147)

[8] M. T. Khan and R. A. Shaik, "High-Throughput and Improved-Convergent Design of Pipelined Adaptive DFE for 5G Communication," in IEEE Transactions on Circuits and Systems II: Express Briefs, vol. 68, no. 2, pp. 652-656, Feb. 2021

[9] M. T. Khan, H. E. Yantır, K. N. Salama and A. M. Eltawil, "Architectural Trade-Off Analysis for Accelerating LSTM Network Using Radix-r OBC Scheme," in IEEE Transactions on Circuits and Systems I: Regular Papers, vol. 70, no. 1, pp. 266-279, Jan. 2023

[10] M. T. Khan and M. A. Alhartomi, "Digit-Serial DA-Based Fixed-Point RNNs: A Unified Approach for Enhancing Architectural Efficiency," in IEEE Transactions on Neural Networks and Learning Systems, vol. 36, no. 5, pp. 8240-8254, May 2025.

[11] Boyang Chen, M. T. Khan, et al., “COMET: Co-Optimization of a CNN Model using Efficient-Hardware OBC Techniques” (arXiv:2510.03516)

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.

Home fee rate and stipend are available for this position.

31 Jan, 2026

Off

Dr Elliot Crowley

Dr Mohd Tasleem Khan

Multi-Agent systems and Data Intelligence

Collaborative Situational Awareness through Geospatial Intelligence

Research Theme

Aim

Develop a framework to integrate and enhance multi-agent sensor data with prior geospatial intelligence to provide robust situational awareness via personalised (AR/VR/Speech UI) and overview (maps, data feeds) interfaces.

Objectives

Develop a decentralised edge computing architecture for geospatial data processing, able to accommodate rapid updates (e.g. sensors), for delivering dynamic situational awareness.
Investigate personalised HCI for different operator roles (e.g. VR/AR interfaces; speech-first hands-free; tangible; egocentric vs allocentric frames of reference).
Assess role-specific HCI adaptations via AI Agents (i.e. the ability for any user to use AI to design their own interface to the system)
Integrate and examine AI techniques for data validation and correction

Description

This research seeks to develop a decentralised edge computing solution to enhance situational awareness in multi-agent systems by integrating live sensor data with geospatial intelligence and real-time modelling.

Remote sensors deployed on robots and in the environment—capturing location, video, audio, and IMU data—will be fused with existing datasets such as LiDAR point clouds, digital surface models to provide personalized situational awareness.

A modular, extensible API will support analyses such as real-time visibility modelling for covert or signal-optimized route planning. AI-driven natural language queries will allow users to interact with the system.

Interfaces will include AR, tablets, and hands-free operation via speech recognition, augmented with LLM-based semantic processing and spatial audio. A key research focus will be integrating AI agents to allow end users—without programming skills—to easily design custom interfaces suited to their needs at any time.

UK 2:1. in computer science / computer systems or MSc in geographical information science.

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.

Home fee rate and stipend available for this position.

31 Jan, 2026

Off

Prof Oliver Lemon

Dr Phil Bartie

Stochastic Contact-Implicit Model Predictive Control and Behavioural Cloning for Versatile Surveillance with Legged Robots

Aim

This PhD project focuses on developing a novel hybrid control framework that combines Model Predictive Control (MPC) and Behavioural Cloning (BC) to enable robust, real-time locomanipulation for navigating and surveillance of complex structures using legged robots. The project aims to train legged robots with motor skills that enable them to perform autonomous surveillance. Specifically, we focus on enabling dynamic motions through MPC, and the versatile behaviour needed to open doors or clear path tasks through BC in quadruped and humanoid robots equipped with manipulators.

Objectives

Design contact-implicit stochastic MPC frameworks that can efficiently compute policy gradients and handle uncertainty in contact events;
Develop a novel diffusion-based learning framework for MPC controllers that can clone dynamic behaviours such as opening doors and path clearance during surveillance operations;
Integrate BL techniques with MPC controllers to execute dynamic surveillance operations autonomously; and
Apply this integrated framework to real-time surveillance on steel and cluttered structures with legged robots.

Description

Model Predictive Control (MPC) has demonstrated remarkable capabilities in enabling agile robotic behaviors—most notably, dynamic maneuvers such as backflips in Boston Dynamics’ Atlas robot. However, conventional MPC methods remain constrained by local optima, limiting their ability to plan complex motion and contact sequences, particularly in cluttered or uncertain environments. These limitations are especially evident in loco-manipulation tasks, where both mobility and interaction with the environment are required.

Moreover, existing whole-body MPC frameworks are largely deterministic, which makes them ill-suited for real-world uncertainties, especially in contact-rich scenarios. On the other hand, behavioural cloning (BC) via diffusion policies has recently shown impressive success in learning the diversity of manipulation behaviors but struggles to scale to more whole-body behaviours where balance and dynamics are critical. Fundamentally, diffusion policies capture and learn the complex distributions present in human behaviours by learning the de-noising process from collected data.

This PhD research will explore a hybrid approach, combining the structure and real-time feasibility of MPC with the flexibility and autonomous capabilities of BC, to enable robust and versatile surveillance on legged robots. Concretely, this project aims to enable legged robots to move around complex environments, requiring them to open doors and remove debris autonomously.

The project builds on advances in robot motor intelligence, differential contact simulation, and model predictive control developed internally in the RoMI lab. It will also advance our current research efforts in Neural Conditioning Probability (NCP) for behavioural cloning.

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.

Home fee rate and stipend are available for this position.

31 Jan, 2026

Off

Dr Carlos Mastalli