About the Project

Today, autonomous vehicle and robotics research is a continuously evolving field and spanning all areas of AI autonomy, including perception, behaviour prediction, motion planning, explainability and safety critical AI systems.

Central to these broader research goals are the ever-present challenges of providing robust on-vehicle perception systems capable of addressing the key research questions relating to vehicle autonomy in terms of "Where am I ?" (localisation), "What is around me?" (scene understanding) and "What is going to happen next?" (motion forecasting) in terms of both scene geometry (3D depth) and semantics (object detection & classification).

This project aims to explore deep learning based computer vision approaches for this task, with a particular focus on developing approaches that can address the long-term challenges of efficiency, in terms of reducing the on-vehicle/robot sensor and processing requirements and hence power usage, whilst also maintaining robust on-task performance under potentially varying environmental conditions.

There are a wide range of topics within this scope such as:

Vision and language models (LLM/VLM) for language-driven perception.
Sensor fusion and multi-modal perception for generalised scene understanding.
Motion forecasting in complex driving scenarios.
Bird’s eye view generation spanning 3D detection, segmentation and occupancy grids.
Multi-agent cooperative perception and planning
Brain Computer Interfacing for robotic control and/or operator/driver monitoring

Supporting Resources

The Department of Computer Science at Durham University hosts well-equipped labs with on-site capabilities comprising vehicle-mounted sensors (LiDAR, radar, camera, GPS/IMU), drone operations, on/off-road robotics including quadruped (dog) and biped (humanoid) platforms, high-precision geo-localisation, BCI/EEG bio-signal data collection, robotic arm manipulation, X-ray security imaging, virtual reality and on-demand wide-area surveillance video feeds.

In addition Durham University hosts the UK regional supercomputer, Bede (128 NVIDIA V100 GPUs) which complements our departmental NVIDIA CUDA Compute Cluster (80+ GPUs up to NVIDIA A100) to cater for the increasing GPU compute demands of modern AI-driven research projects.

The department itself is based in the newly built Mathematical and Computer Sciences building - a £42 million, 9,160m2 state of the art teaching, learning and research facility, located on the University's Upper Mountjoy Campus co-located with the Department of Mathematics. All PhD students are allocated desk and/or lab space to support their research work and PhD student research projects are additionally supported by an annual equipment/travel allowance spanning the duration of the PhD study period.

Supervision

You will be supervised by Prof. Toby Breckon, Professor of Computer Vision and Image Processing in both the Department of Computer Science and Department of Engineering, and Head of Visual Computing (Computer Science) at Durham University in collaboration with one or more staff from the VIViD research group.

Entry Requirements

First class honours (or high 2.1 at undergraduate or equivalent Masters) degree in a Engineering, Physics, Maths or Computer Science based subject (internationally: GPA 3.3+ or equiv).
Strong understanding of computing/engineering applications and mathematical problem solving.
Knowledge of modern programming languages (ideally including one of Python, C++/C or Rust).
Excellent written and spoken communication skills in English, meet the Durham University English language requirements.
Prior experience in research, industry and/or AI based research topics is beneficial but not essential.

How to Apply

Please send an email with your CV, degree transcripts and any supporting documents to Professor Toby Breckon toby.breckon@durham.ac.uk for an initial pre-application discussion.

Funding Notes

This is primarily a self-funded PhD position and applications are welcome all year round with flexibility on start date.