Cookies

This site requires the use of cookies as defined by our Terms and Conditions.  We have provided a detailed description of how cookies work and are used on the site.  To accept cookies, please click the "Accept Cookies" button.

Due to essential maintenance, this site will be unavailable from 8am on Wednesday 26th June for approximately 1 hour

View All Vacancies

PhD Studentship - Deep Learning in Graph Domains for Sensorised Environments

Engineering and Applied Science - Studentships

Location:  Aston University Main Campus
Basis:  Full Time
Closing Date:  23.59 hours BST on Friday 12 July 2019
Reference:  R190198

PhD Studentship (3.5 years)

Supervisors: Luis J. Manso, George Vogiatzis

Key words: deep learning, graph neural networks, smart environments, robotics

Applications are invited for a studentship in the Aston Institute of Urban Technology and the Environment (ASTUTE), funded by the School of Engineering and Applied Science. The successful applicant will join a cohort of graduate students working on projects across the broader Smart Cities field, and as part of the PhD will receive training and experience in collaborative research, relevant to industry and Smart City planners. The studentship is offered in collaboration with Ortelio Ltd.

The position is available to start in October 2019.

Financial Support

This studentship includes a fee bursary to cover the home/EU fees rate, plus a maintenance allowance of £15,009 in 2019/20 (subject to eligibility).

Overseas Applicants

Applicants from outside the EU may apply for this studentship but will need to pay the difference between the ‘Home/EU’ and the ‘Overseas’ tuition fees, currently this is £12,573 in 2019/20.

As part of the application you will be required to confirm that you have access to this additional funding.

Project Description

The future will bring an increase in the presence of robots in smart urban environments, especially in public buildings and their surroundings. Robots and sensorised environments will cooperate to assist people and perform automated monitoring tasks to enhance their experience in the building and ensuring that all facilities are working as expected.

These environments are characterised by a large number of sensors generating vast amounts of complex and interrelated data which is naturally expressed as graphs. For instance, nodes can be people, rooms, sensors or appliances. Nodes have multiple properties which depend on their type, and edges represent relationships between them. The following figure depicts a simplified version of the kind of data structure that these environments can generate.


The predominant Machine Learning (ML) approach to deal with graph structures is two-stage. In the first stage, input data is transformed into vectors so that traditional techniques can be used. In the second stage the generated vectors are processed to learn and extract conclusions from 

the data, e.g., classifications, regressions, anomaly detection. While this works for simple tasks, information is generally lost in the graph-to-vector conversion, and not all the input information is available for the second step, which is the one actually performing the reasoning. Unfortunately, using ML techniques natively working with graph-representations is a much less explored approach. This PhD will focus on enabling robots and environment sensors to share a common graph-like world model representation and to perform reasoning tasks using machine learning techniques specially designed to work with graph structures. Cloud computing technologies will be used to overcome scalability issues. The indirect applications of the research outputs are wide, as the techniques developed will be used for general purposes. However, the PhD will be heavily focused on smart environments, having the following main applications:

•    Classification and prediction of user’s intentions: it can in turn be applied to managing building settings such as lighting conditions or heating systems.

•    Activity recognition: it will help understand users’ intentions and feelings, and to detect suspicious and dangerous activities.

•    Learning of domain-specific task planning heuristics and useful actions: to enable robots compute plans efficiently in complex settings.

Person Specification

The successful applicant should have a first class or upper second class honours degree or equivalent qualification in Computer Science, Electronic Engineering, Mathematics, Physics or similar.  Preferred skill requirements include knowledge/experience of programming, deep learning, mathematics.

We would particularly like to encourage applications from women seeking to progress their academic careers. Aston University is committed to the principles of the Athena SWAN Charter, recognised recently by a prestigious Silver Award to EAS, and we pride ourselves on our vibrant, friendly and supportive working environment and family atmosphere.

Contact information

For formal enquiries about this project contact Luis J. Manso by email at l.manso@aston.ac.uk. 

Submitting an application

Details of how to submit your application, and the necessary supporting documents can be found here.  

The application must be accompanied by a “research proposal” statement. An original proposal is not required as the initial scope of the project has been defined, however candidates should take this opportunity to explain how their knowledge and experience will benefit the project and should demonstrate their familiarity with some relevant research literature.

If you require further information about the application process please contact the Postgraduate Admissions team at seasres@aston.ac.uk 

Email details to a friend

Further particulars and application forms are available in alternative formats on request i.e. large print, Braille, tape or CD Rom.

If you have any questions, please do not hesitate to contact HR via recruitment@aston.ac.uk

 

 

 


Login

Login

Forgotten Details

Register