PhD projects 2010

The project suggestions here give you an idea of the possibilities for research projects with SICSA partners. However, there are NOT the only opportunities - you should also look at partner websites, understand the research that is going on and make your own proposal for a research project.

Abertay University

Abertay has world leading courses in Computer Games Technology, is the UK’s first Centre for Excellence in Computer Games Education and has recently been awarded £5 million to support the creation of new games and bring them to market.




Biological metaphors for managing speckled computing networks



Speckled computing is a model of spatially distributed computation based on specks, minute, autonomous semiconductor grains that can sense and compute locally and communicate wirelessly. Thousands of specks may be scattered in an environment to provide a low-cost, efficient distributed sensing network. However, managing the SpeckNet efficiently, attracts several challenges including coping with node failure, optimising performance with power-limited devices and ensuring propagation of information across the network.

Biology has many examples of evolved systems that demonstrate robustness to failure, information propagation in harsh environments and energy-efficient solutions to resource flow. This project will investigate the use of existing computational models of biological systems, specifically fungal colonies and the immune system, as paradigms for dynamic management of speckled computing networks. Based at the University of Abertay, this project combines expertise in fungal colonies (Abertay), artificial immune systems (Edinburgh Napier) and speckled computing (Edinburgh).





Enhancing interactive game play through evolutionary computation



In many genres of computer games the increased complexity of interactions between the game player and the non-playing characters, and between these characters themselves, has led to the necessity for the computer controlled characters to be able to learn and adapt in ways that appear more natural and human-like than has been possible with standard game artificial intelligence (AI) techniques such as scripting, rule-based decision making and finite state machines. Therefore, much research has been undertaken in the development of autonomous, intelligent agents for computer games.

Although various AI techniques, such as Artificial Neural Networks, Genetic Algorithms, and Fuzzy Logic have been used to a limited extent, to give the appearance of more believable, human-like behaviours, they have not yet been used together in a structured way to create a truly autonomous, intelligent agent. This project will investigate the possibility of combining AI techniques to produce an autonomous intelligent agent that is able to learn, adapt and plan a course of actions based on interactions with other game characters, players and its environment.

Glasgow University

Parallelising model checking using FPGAs (Dr A. Miller, and Dr W .Vanderbauwhede)

Model checking is a technique used to verify properties of concurrent, communicating systems. An abstract model of a system is created and an automatic tool - a model checker - is used to exhaustively search an associated graph (the state-space) to check the behaviour. One of the major problems with model checking however is that state-spaces can become very large. This is known as the state-space explosion problem, and a rich area for research is tackling this problem.

FPGAs are a type of integrated circuit that can be programmed. There is a growing interest in the use of FPGAs for computational tasks as they are potentially much faster than a microprocessor, with a fraction of the power consumption.

This project will investigate the use of FPGAs to alleviate the state-space explosion problem, by allowing parts of the model checking process to be handled in parallel. Different model checkers use different search algorithms and allow for a variety of different types of behaviour to be investigated (e.g. non-deterministic, probabilistic, real time). The aim of this project is to determine which model checkers, and which aspects of their implementation succumb to parallel techniques.

Investigations into Algorithmic Information Theory. Dr Paul Cockshott and Dr. Lewis Mackenzie (Glasgow), Dr. Greg Michaelson (Heriot-Watt)

The original definition of entropy by Boltzmann is anthropomorphic because it does not base itself on a fundamental law of physics but on human ignorance about the possible microstates of a system. A similar objection can be levelled against Chaitin’s definition of algorithmic entropy as being conditional on the definition of the UTM used for defining the shortest algorithm to compute a number. The Boltzmann definition has been developed by physics since then because quantum mechanics gives a lower bound to the microstates that a system can occupy [2] – these being anthropic in the Boltzmann case. The research problem is to investigate whether the computational definition of entropy can be appropriately renormalized to remove the anthropic decision on what TM implementation is to be used. The research project could involve large parallel computing task to approximate the first n digits of the omega number. The department can call on substantial parallel computing resources through SCOTGRID and via EPCC. Omega is the probability that a TM will halt on a randomly given input tape. For background see [1]

[1] Exact approximations of Omega numbers. CS Calude, MJ Dinneen - Int Journal of Bifurcation & Chaos, 2007
[2] Von Neumann, John (1955). Mathematische Grundlagen der Quantenmechanik (Mathematical Foundations of Quantum Mechanics). Berlin: Springer. ISBN 3540592075

More project suggestions from Glasgow.

Heriot-Watt University

Simple neural network models (Dr. P. Frisco, Heriot-Watt University and Prof. L. Smith, University of Stirling)

The aim of this research project is to find the laws linking static properties of simple neurons and models of neural networks to the dynamic properties of these networks. The overall aim of this area is understanding the nature of neural processing: we will therefore assess these networks against both what they can achieve in an engineering sense, and how their properties compare to those of specific real neural systems.

Architecture transparent control of parallelism through parallel architecture models. (Hans-Wolfgang Loidl. Greg Michaelson, Phil Trinder; Kevin Hammond, St Andrews)

Parallel architectures are increasingly heterogeneous and hierarchical. That is they contain a range of processor types: cores, CPUs, GPUs, FPGAs etc; moreover these processors communicate over an hierarchy of communications networks, e.g. cores in a cluster of multi-cores communicate by shared memory or by some switch/LAN.Rather than expect the programmer to manage the intricacies of these architectures, and revise the program for each new architecture, we would like to automatically adapt the program to the specific target architecture. This could be done as part of an implementation of an abstract machine, providing a model of semi-explicit parallelism, or by implementing algorithmic skeletons, capturing common patterns of parallelism. Both directions are currently explored in different implementations.

The aim of the project is to identify a suitable parallel architecture model, and to use its information to automatically coordinate the parallel execution on a range of parallel architectures. The objectives of the project are as follows.

• To research parallel architecture models and identify one at an appropriate level of abstraction.
• To investigate using the architecture model to automatically control parallelism within an abstract machine for semi-explicit parallelism or within parameterised algorithmic skeletons.
• To evaluate the results on a range of alternate target architectures, including multi-cores and clusters of multi-cores.

Robert Gordon University

Project suggestions from Robert Gordon University.

St Andrews University

Mobile sensor swarms (Prof Simon Dobson)

Environmental sensing often involves sampling data across a wide area. There are basically three ways to do this: observe from a distance, deploy a large number of static sensors, or deploy a smaller number of mobile sensors. The first can be inaccurate and expensive, the second intrusive. Mobile "swarms" of sensors potentially offer a way forward, but for them to be practical we need to solve some complex problems. How is the swarm co-ordinated? How can we have confidence that the data collected is scientifically valid? How can we re-purpose the swarm over time? This work will look at swarm sensing both theoretically and in practice, to develop the techniques needed to perform large-scale, long-lived observations.

Mapping and Analysing Information in Online Social Networks and Blogs (Dr Mirco Musolesi)

The aim of this investigation is to study and model dynamic processes in online social networks and in the blogosphere. Two types of projects are possible: 1) one more theoretical focussing on the design of novel diffusion models of information in networks and the analysis of real datasets from online social networks such as Facebook and Twitter; 2) one more practical focussing on the implementation of systems for mapping and analysing information diffusion in networks in real-time.

More project suggestions from St Andrews.

Stirling University

Scalable Adaptive Multicasting (Dr. M. Kolberg)

This research will focus on developing a scalable and adaptive approach for the combination of application layer multicast, native multicast, and multicast tunnelling. The concept of scalable adaptive multicast includes both scaling properties and adaptability properties. Scalability is intended to include diverse properties such as large group sizes, large numbers of small groups, and very high or low rates of group membership change. Adaptability includes the use of different control mechanisms for different multicast trees.

Rigorous Decision Support for Healthcare (Prof. K. J. Turner)

This research will design a generic and rigorous methodology for creating clinical decision support systems. The approach will be evaluated using chronic heart disease as an exemplar condition. The focus is on developing new techniques for designing decision support, and will complement existing guidelines, models, methods, formats and tools. The plan is to concentrate on two key aspects of decision support design that need improvement: abstractness and analysis. A more detailed description of this topic is available online.