See also
Intelligent Systems Research Group - PhD Research Projects
PhD research projects are currently available in the following areas. Some projects listed may also be suitable for the MSc by Research – please contact the member of academic staff involved if you are interested in the MSc by Research. Please note that funding opportunities, if available, are advertised on our Funding page.
- A new crossover technique in Cartesian Genetic Programming
- Application of CGP to control problems
- Application of CGP to function optimization
- Application of CGP to the object identification in images
- Applying evolutionary techniques to animal species identification
- Augmented video for medical applications
- Automated recognition of vehicle types
- Automatic face description in computational photography and videography
- Automatic recognition of manmade versus artificial sounds using CGP
- Biologically-inspired metamorphic architectures for failure tolerance in flying robots
- Computational neuroscience: neural signal processing & neural computing
- Developmental dynamics for Network-on-Chip systems
- Evolution of music using CGP
- Evolutionary algorithms in the detection of Parkinson's Disease
- Evolutionary mapping of algorithms to arrays of processors
- Hybrid description methods for human faces
- Image analysis of mammograms towards the classification of breast cancer
- Immune-inspired robotic systems
- Investigations into autonomous adaptable hardware
- Low Power Design of FPGA Systems
- Neural-endocrine behaviour based robotics
- Self-regulation of robotic systems using artificial immune systems
- Simulation of complex systems on reconfigurable hardware
- Towards the diagnosis of Alzheimer's disease
- Using evolutionary computation to design the next generation of nano-CMOS systems
A new crossover technique in Cartesian Genetic Programming
Dr Janet Clegg, email: jc@ohm.york.ac.uk
This project would look at testing a new crossover technique in Cartesian Genetic Programming. The new crossover technique has been tested on simple regression problems and has proved extremely successful. It now needs testing on other types of problems; for example circuit design problems. The project would then look at applying the technique to medical diagnosis applications. In particular, the diagnosis of Parkinson's disease and the measurement of body water content. Should this project be successful, it will potentially have a large impact on many people's lives.
Application of CGP to control problems
Dr Julian Miller, email: jfm7@ohm.york.ac.uk
In a recent excellent final year project, CGP was used to evolve a helicopter controller with surprising and excellent results. This year a student has just been awarded a PhD for his work on using CGP for automatic control during sensor failure. There are many problems in control that can be tackled with CGP and it would be fascinating to not only solve them but examine how the evolved programs work. This topic would be supervised jointly with Tim Clarke.
Application of CGP to function optimization
Dr Julian Miller, email: jfm7@ohm.york.ac.uk
Function optimization is an extremely important topic. Many real-world problems can be formulated as finding a series of numbers that make a complex function an optimum. A recent paper by James Walker and Julian Miller presented a way of using CGP to solve conventional optimization problems. There are a number of other ways that CGP could be used for such problems. This subject would make a very good PhD topic as it is a very novel idea with a great deal of potential and already has early promising results.
Application of CGP to the object identification in images
Dr Julian Miller, email: jfm7@ohm.york.ac.uk
Identifying whether images contain certain objects and features is something human beings do very well. However it is a very challenging problem to write computer programs that can do this. CGP offers great potential in this area as it could evolve very novel ways of doing this. Dr. Miller has formulated some feasible and fascinating approaches to this problem. Research on this topic would have very wide application. In fact this might be several PhDs! Since different PhDs could be specialized on different types of images (medical, agricultural,…)
Applying evolutionary techniques to animal species identification
Dr Janet Clegg, email: jc@ohm.york.ac.uk
This project will investigate the identification of animal species using evolutionary techniques. The identification of species using bioacoustics and image processing together with an artificial neural network is already available and tested in work by Dr Dave Chesmore. The new idea would be to substitute the artificial neural network within the process with an evolutionary technique. Evolutionary techniques have recently proved extremely successful in solving large complex problems. It is hoped that, by substituting the evolutionary element into the species identification process, its performance in correctly identifying species will be improved.
Augmented video for medical applications
Prof John Robinson, email: jar11@ohm.york.ac.uk
The group has long experience in the development of "Video Augmented Environments", where a camera watches a working surface on which users move objects or draw or gesture, and in response the system projects graphics and makes sounds. This project is concerned with projections onto the human body. These may be renderings of anatomical information (e.g. bones, blood flow) which change as the person moves. Such a system could be used in teaching of medical examination techniques. Other applications where data are not projected onto the body, but instead an augmented view is seen on a screen, will be investigated. In all cases the system must process input video in real time and render registered (aligned) overlays.
Automated recognition of vehicle types
Prof John Robinson, email: jar11@ohm.york.ac.uk
In collaboration with two industry partners, this project will investigate the automated recognition of car and truck makes and models. We will apply feature-based and appearance-based classifiers to this problem, exploiting known structures in the scene (e.g. the planarity of licence plates).
Automatic face description in computational photography and videography
Prof John Robinson, email: jar11@ohm.york.ac.uk
The group has recently filed a patent on a method for estimating the attributes of faces detected in pictures. These attributes include the location and orientation and facial "landmarks", such as the eyes, nose tip and chin tip, along with the age, sex, race and facial expression of the person. A visualization of the method working frame-by-frame on a video can be seen at http://www.elec.york.ac.uk/visual/jar11/kbdemo2.avi. This project's aim is to apply these methods to computational photography, linking with other work in the group on shake removal and high-dynamic-range imaging. The project will use information about facial position, pose, eye gaze and expression to control enhancement and restoration algorithms. When applied online within the camera, this will aid in capturing stills of highly-dynamic scenes of people (potentially with application to surveillance as well as photography). When applied offline, it will assist with automatic portrait touch-up and the restoration of old photos and films.
Automatic recognition of manmade versus artificial sounds using CGP
Dr Julian Miller, email: jfm7@ohm.york.ac.uk
Humans beings are very good at distinguishing manmade sounds from natural sounds (i.e. the sound of a real piano from the sound of an electronic piano). This project is about evolving programs (or circuits) that can solve this problem using CGP. This would have a number of uses. It would be useful also in devising circuits or programs that can produce much more natural sounds. Such a project would be supervised in cooperation with Prof David Howard.
Biologically-inspired metamorphic architectures for failure tolerance in flying robots
Dr Andy Pomfret, email: ajp109@ohm.york.ac.uk & Prof Jon Timmis, email: jt517@ohm.york.ac.uk
Biological systems can display impressive adaptivity to increase their own 'dependability' and ensure survival. Initial work at the University of York has identified structural adaptivity as being key to the implementation of generic, failure-tolerant control systems. Therefore the aim of this project will be to investigate evolvable, structurally adaptive systems for the control of a UAV, driven by predictions of future failures, such that performance can be maintained in the face of unplanned failures. The project will have access to several flying robots, including a large Hexacopter 6-rotor UAV and a number of smaller quadrotor craft. The new, double-height Robot Lab on the Heslington East campus will also be available for conducting experiments.
Computational neuroscience: neural signal processing & neural computing
Dr David Halliday, email: dh20@ohm.york.ac.uk
The human brain is one of the most complex systems known to man. It consists of vast numbers of interacting units (called neurones) that communicate with each other using sequences of pulses (called spike trains). The highest levels of processing in the human brain occur in the neocortex. One mm3 in the neocortex contains around 100,00 cells, 4 km of wiring (called axons) 500m of cell body outgrowths (called dendrites) and up to 109 connections (called synpases) between cells. Much is known about the organisation and function of the human brain. It is possible to study the brain experimentally and to simulate the behaviour of neurones using digital computers. A characteristic feature of all neural systems is that they are intrinsically noisy. Any signal processing techniques used to study neural communications must take this into account.
Such a complex system provides plenty of scope for research projects in Computational Neuroscience. Important areas are the analysis and modelling of neural systems. There is a need to develop multivariate statistical signal processing methods that can be applied to neurophysiological data sets to extract information about the structure, function and operation of the nervous system. This can be supported by a programme of computational modelling using computer models to simulate neural behaviour. These models can be detailed biophysical models or more general purpose bio-inspired models. Simulation can be done in software or hardware (using FPGA technology). Precise project details will depend on your interests.
Developmental dynamics for Network-on-Chip systems
Dr Gianluca Tempesti, email: gt512@ohm.york.ac.uk
The Network-on-Chip (NoC) paradigm, a subject of intense research in both the academic and the industrial worlds, is based on the parallel execution of program threads on many relatively simple processing elements. The mapping of the threads to the processing elements is, conventionally, a static process, based on performance estimates carried out at compile time. The aim of this project, which builds on past work on the design of custom processors, is to investigate how some of the mechanisms that control the development of multi-cellular organisms (growth, cell death, cicatrization, ...) can be applied to networks of processing elements to optimize their computational performance (for example, by replicating overloaded processors or eliminating idle ones at runtime).
Evolution of music using CGP
Dr Julian Miller, email: jfm7@ohm.york.ac.uk
Already artificial evolution has been used to evolve music and art. In fact CGP itself has been used to evolve art. It would be fascinating to apply CGP to the evolution of music, this could be for automatic improvisation, and a whole range of exciting creative areas. This project would be likely to be supervised jointly with Dr Andy Hunt.
Evolutionary algorithms in the detection of Parkinson's Disease
Dr Steve Smith, email: sls@ohm.york.ac.uk
The Department is currently undertaking clinical trials of a new data acquisition system to aid diagnosis and monitoring of Parkinson's disease. The research employs an evolutionary algorithm (based on Cartesian Genetic Programming) in the analysis of the data resulting from this trial with view to evaluating the diagnostic value of the technique. The work is pioneering and therefore at an exciting stage and will involve close collaboration with two major Hospitals in the UK for discussions with clinical staff at the most senior level. The project is concerned with using evolutionary algorithms to discover new characteristics of Parkinson's patients that will enable better monitoring of their condition.
Evolutionary mapping of algorithms to arrays of processors
Dr Gianluca Tempesti, email: gt512@ohm.york.ac.uk
The Network-on-Chip (NoC) paradigm is a subject of intense research in both the academic and the industrial worlds. Based on the parallel execution of program threads on many relatively simple processing elements, this approach suffers however from a crucial weakness: the difficulty of finding an efficient way to map computation to the array of processors. Building on past work on the design of custom processors, this project seeks to investigate the use of evolutionary algorithms to find acceptable solutions for this NP-complete problem. The project will take into account the optimization of performance at the same time in single processors and in the complete network, by analyzing both the single threads and their communication patterns.
Hybrid description methods for human faces
Prof John Robinson, email: jar11@ohm.york.ac.uk
The group has pioneered the application of conditional probability density estimation to describing human faces. Our methods allow a system to estimate the age, sex, race and facial expression of people in a real-time video sequence. Accuracy is acceptable for some uses but not for demanding security applications where lighting may be poor and faces partly occluded. To tackle these difficult cases, we will combine our methods with new approaches in face landmark extraction, motion analysis and cascade classification. This project will explore these broad alternatives before focusing on promising avenues that will result in high-accuracy, high-speed robust face description.
Image analysis of mammograms towards the classification of breast cancer
Dr Steve Smith, email: sls@ohm.york.ac.uk
Breast cancer claims many thousands of lives each year in the United Kingdom alone. There is an increasingly severe shortage of radiologists in the UK to diagnose the onset of this disease from the breast-screening X-ray mammograms. This problem is compounded by the disease-preventative need to extend the screening to a wider age group, and to double the number of images taken per session. There is a very real need for computer-aided detection (CAD) equipment in the breast screening units, which would support the clinicians in detecting breast cancer at the earliest possible stage. This project work is an investigation into ways of very accurately detecting residues of calcium, known as microcalcifications, which are known in certain types of breast cancer to be a very early, pre-tumourous, indicator of disease progression. Microcalcifications are quite often indistinct in the grey-scale mammographic images, having low contrast against the highly variable image structures produced by the surrounding breast tissue. However, optimal detection of microcalcifications, avoiding both false-negative and false-positive detections, is a crucial operation in providing clinically acceptable CAD support equipment, capable of appropriately marking regions of suspicion on the mammograms to aid the radiologists in their diagnosis. This investigation into optimal detection of individual and grouped microcalcifications will involve the use of evolutionary algorithms such as Cartesian genetic programming on the raw pixel values.
Immune-inspired robotic systems
Prof Jon Timmis, email: jt517@ohm.york.ac.uk & Prof Andy Tyrrell, email: amt@ohm.york.ac.uk
On-going work in the department is concerned with abstracting metaphors from the natural immune system into a variety of engineering applications. As part of the process, we make use of a modelling tool called Stochastic-pi calculus, a tool designed for the modelling of concurrent systems. We have been working on the development of models of the tunable activation thresholds in T-cells as a means to developing systems that exhibit tunable, homeostatic behaviour. This research would extend our current investigations using Stochastic-pi calculus for the modelling of tunable activation thresholds in T-cells and the creation of novel immune inspired systems for the deployment in swarm robotic systems.
Investigations into autonomous adaptable hardware
Prof Andy Tyrrell, email: amt@ohm.york.ac.uk
For systems to become truly autonomous we must develop methods that will allow hardware devices/systems to adaptable in an intelligent and appropriate way to environmental changes and component failures. This project will consider the use evolution to automatically design hardware capable of self-adaptation. The focus will be on the hardware’s ability to demonstrate adaptability and fault-tolerance in particular. To achieve these goals it will be interesting to understand how artificial evolution can utilize novel properties of physical devices to achieve large scale and complex solutions to a number of computational problems.
Low Power Design of FPGA Systems
Prof Andy Tyrrell, email: amt@ohm.york.ac.uk & Dr Gianluca Tempesti, email: gt512@ohm.york.ac.uk
Field programmable Gate Arrays are becoming the device of choice for many (most) digital systems. They offer the flexibility of software with the performance of hardware. The downside of using FPGAs is their propensity to be high on power. Through analyzing and investigating novel reconfiguration and other techniques, it should be possible to produce innovative mitigation methods for the design of lower power FPGAs designs. This project will focus on the use of bio-inspired methods to add these low power characteristics to FPGA designs.
Neural-endocrine behaviour based robotics
Prof Jon Timmis, email: jt517@ohm.york.ac.uk
On-going work in the department is developing novel approaches to long-term autonomy through the development of a novel neural-endocrine inspired behaviour control system. This research would investigate novel approaches to the neural-endocrine system, and aim to deploy a control system on a robotic for long periods of time (days) on a robotic system either land-based, or water based.
Self-regulation of robotic systems using artificial immune systems
Prof Jon Timmis, email: jt517@ohm.york.ac.uk & Prof Andy Tyrrell, email: amt@ohm.york.ac.uk
In order to maintain homeostasis, there exist many systems within a biological organism that, through their interactions, give rise to stability. These interactions are widely acknowledged as operating between the immune, neural and endocrine systems. However, homeostasis also occurs individually within each one of these systems. Endowing engineered systems with such homeostatic properties would have significant benefits. The aim of the work is to exploit immunological properties to produce an integrated hardware and software architecture that can be employed within an embedded dynamic system (robots). This combination will endow large-scale pervasive systems (robots) with the ability to self-regulate their physical and operational state within highly dynamic environments.
Simulation of complex systems on reconfigurable hardware
Prof Andy Tyrrell, email: amt@ohm.york.ac.uk & Prof Jon Timmis, email: jt517@ohm.york.ac.uk
The research will investigate advanced FPGA architectures that better support complex system simulations, by considering the practical implications of implementing complex system models on a hardware platform. Initially FPGA requirements will be considered. The issues that will be considered include: what hardware structures are needed to support highly parallel interacting mobile processes? What processes best provide support for dynamic reconfiguration? How best can hardware support the communication issues (both between agents, and in extracting simulation data in real time for analysis)? Solutions will be prototyped, leading to a specification for a "Complex System" ASIC.
Towards the diagnosis of Alzheimer's disease
Dr Steve Smith, email: sls@ohm.york.ac.uk
Recent research in the Department has identified the benefits of using evolutionary algorithms such as Cartesian Genetic Programming (CGP) for the analysis of patient responses to neuropsychological assessment. This particular project will focus on the objective assessment of spatial ability which can be of use in diagnosis of neurodegenerative conditions such as Alzheimer's disease. Current work is in close collaboration with consultants from a major hospital in the UK and will involve clinical trials, the data from which will be used to develop the evolutionary algorithms.
Using evolutionary computation to design the next generation of nano-CMOS systems
Prof Andy Tyrrell, email: amt@ohm.york.ac.uk
It is widely recognised that variability in electronic device characteristics and the need to introduce novel device architectures both represent major challenges to scaling and integration for present and next generation nano-CMOS transistors and circuits. The rapid increase in intrinsic parameter fluctuations, as devices get smaller, stemming from the fundamental discreteness of charge and matter and their statistical impact on device behaviour is a major source of device variability. The intrinsic parameter fluctuations are fundamental, truly stochastic and cannot be eliminated by tighter process control. The work in this research project, will study the impact of next generation technologies, and related parameter fluctuations, on the design of digital circuits using evolutionary computation. In particular we will attempt to determine answers to the following fundamental questions: How can evolutionary techniques be used within device models to alleviate parameter fluctuation problems? How can parameter variation datasets be best used by evolutionary techniques to improve system performance? How can evolutionary techniques be used to limit the effects of parameter variations?
Postgraduate Admissions Tutor: Dr. Steve Smith
Tel: (+44) 01904 324485