PhD in Computing Science - Computational Intelligence for algorithmic support of teams in data-intensive contexts

A key challenge in data science and AI is how best to bring human insight, sophisticated analysis algorithms and large amounts of time-critical data. Many environments such as emergency or military services present highly complex situations in which crews need to make potentially critical decisions relating to platform command and control, threat evaluation and interactions with other elements of the service. To make those decisions effectively they must be able to assimilate large quantities of information regarding their surroundings.

Thales is involved in systems design for vehicles with rapidly increasing levels of sensor and information systems, designed to give crew improved situation awareness. Such systems require Sensor Fusion to combine data from multiple sources. They apply Information Management and Information Exploitation (IM/IX) to filter raw data into representations that are useful to the crew. Based on this information vehicle crews make correct and timely decisions.

This Ph.D. studentship between the University of Glasgow, Quantic and Thales UK will explore the topic of Human-Machine teaming in situations with time-sensitive decision making, cognitive and data overload and significant information uncertainty. The project will explore the use of machine learning and other AI techniques to help fuse multi-source information and provide decision support for a team of humans. The work is likely to give the candidate experience in evaluating the performance of an AI agent, developing Explainable AI techniques, and checking for bias in decision making will be key for the agent to gain trust from human members of the team.

The work will be based around a number of mini-projects, together with colleagues at Thales, where adaptations of specific challenges associated with their ongoing projects are linked into the ongoing fundamental research in the Ph.D. The student will be primarily based at the University, but is expected to spend time at the Thales offices in Glasgow appropriate to the project work scope, and which could include data gathering and experimentation. The project context below describes the broad area of interest. It would be expected that the student would carve out a specific task in this area together with supervisors from the University and Thales.

Project Context:

Thales and the University are already working together in a project with QinetiQ and other industrial partners relating to how best to design information and control systems for vehicles in challenging urban environments. The student will have the opportunity to work with topics such as machine learning, sensor fusion algorithms and human-computer interaction.

Thales will provide access to vehicle subsystems and data gathering tool sets. The Information, Data and Analysis Section of the University of Glasgow has wide-ranging skills in Machine learning, Computational Interaction Design and Vision systems, and is also part of the Quantum Imaging Hub, all of which are highly relevant for this task, so the student will have a range of skills and approaches to support them within the University. The work has Computational Interaction at its heart: Computational interaction applies computational thinking (abstraction, automation, analysis) to explain and enhance interaction between a user and a system. It is underpinned by modelling which admits formal reasoning, and which is amenable to computational approaches, and draws on insight from machine learning, signal processing, information theory, optimisation, Bayesian inference, control theory and formal modelling. This is an area the University of Glasgow has been leading development in, so the research brings together leading edge machine learning and developments in computational interaction to address challenging problems in human/data/algorithm interaction.

Research into AI/human teams typically involves analysis of the team roles and the design and development of AI agents which will act as virtual team members. AI agents will track team member attention and behaviour to monitor whether they are currently performing appropriately, or whether data or cognitive overload is kicking in, and decide whether they need extra support. The type of shared control support provided by the agents could include:

a) Dynamically changing degrees of freedom in sensing and control for different participants depending on their load, reducing the data flow to a level that is achievable in the current context.

b) Dynamic approaches to transitioning between human and machine control in a team setting, such that algorithmic semi-autonomous agents take over certain tasks, to permit the human team members to focus on critical aspects.

c) Use of Machine learning tools to analyse variability and uncertainty in information flows, with the goal of using this information to prioritise the most important and reliable information for decision makers.

This role will require Security Check (SC) Clearance. If not currently held it is expected that the post holder will undergo SC Clearance, please visit the UKSV website for further guidance.

Please refer to the following website for details on how to apply:

http://www.gla.ac.uk/research/opportunities/howtoapplyforaresearchdegree/

Professor Roderick Murray-Smith http://www.dcs.gla.ac.uk/~rod/

Dr Fani Deligianni http://fdeligianni.site/