Single and Cooperative Imaging based GPS Denied Localisation Solution for Autonomous Platforms

Full Studentship including Industrial Placement, Duration: 3 years

With our esteemed partner, we are looking for a well-qualified and motivated PhD student to conduct the following fascinating research project:

Cranfield University and DSTL have defined a program of research entitled "Single and Cooperative Imaging based GPS Denied Localisation Solution for Autonomous Platforms". The PhD has the ambition to tackle the fundamental problem of providing a reliable and efficient mean of Camera based and GPS denied navigation solutions, more convenient to use, not prone to jamming or bad reception and can be more precise than IMU in the long run. However, clouds and other obscurants can be an issue as can illumination levels at night, especially where a platform wishes to remain stealthy. The following paragraphs build the case for the solution to be investigated in the PhD.

Visual Odometry (VO) is a camera based navigation technique, used in a wide variety of robotic applications that estimates the motion of a vehicle from the differences between images taken at regular intervals. Images can be monocular or stereo. Unlike similar algorithms such as Simultaneous Localisation and Mapping (SLAM), VO focuses on relative position rather than mapping the terrain.

Passive Thermal Imagery (TI) which detects an objects emittance rather than its reflection is able to be used at night and is unaffected by illumination changes. Thermal cameras are also becoming more ubiquitous on vehicle platforms and becoming more affordable as technology advances. There are challenges associated with TI such as low spatial resolution, history effects, variations in temperature, and low signal-to-noise ratios: New image processing methods would need to be created for this type of sensors if used a VO navigation solution.

In terms of navigation algorithms to develop for this PhD study they would range from using classical optimal and robust filtering able to provide a tool of fusion integration of different sensing modalities (cameras and Inertial sensors) to the use of new techniques based on Deep Learning strategies based on Convolutional Neural Networks extendable to deal with this kind of heterogeneous data modalities.

Entry Requirements

Applicants must be UK nationals. Applicants should hold at least a Bachelor Degree of either First Class or hold a Masters Degree (MSc/MRes) in Electrical/Computer Science/Applied Mathematics or any other relevant discipline.

Candidates should have a good basis in mathematics with good scientific programming skills (Matlab, C++). Experience in Navigation systems, Computer Vision and Machine Learning is an important asset. Good interpersonal and communication (oral and written in English) skills are also required.

How to apply

A CV and the name and address of two referees should be sent to Prof Aouf at [Email Address Removed].