RAE2001 logo

Submissions

 
 

RA5a: Structure,environment and staffing policy

Strategy
Computer Science resides within the School of Computing and Mathematics and is the dominant component of the School (approximately 90% of the academic establishment). The key elements of the School’s research strategy are
§ the School’s research is to be mainly applied in nature but with elements of fundamental research
§ the School’s research is to underpin its teaching programmes
§ to formally recognise groupings of research staff in specific domains
§ for research to be interdisciplinary and multidisciplinary whenever possible
§ for School funded activities to be focussed on members of staff at the start of their research active careers.
These elements came out of a major review of the School’s research policy in 1999 following a University wide review of research completed in 1998. The School’s revised research policy identified nineteen specific issues to be addressed, along with time scales for each issue, which fell into one of three areas; namely, culture/environment, funding and roles/responsibilities. A few of the specific issues identified are outlined below.
Key elements of the research strategy outlined in 1996 remain; for example, clearly identified groupings, research to be mainly applied in nature. What has changed are some of the groupings, the University environment (e.g. the introduction of University sabbaticals), a more explicit strategy in detailing how the policy is to be implemented and physical changes to ensure research active staff and students are housed together to help a research culture flourish.
The revised policy covers both RAE categorised research and research that falls outside the RAE exercise. A major outcome of the research review was the allocation of over 500 square metres of contiguous space in one building for the School’s research active staff, students, equipment and meeting room. This space houses the research activities described in this RAE submission as well as the externally funded TLTP/FDTL projects, work on learning technologies and related research activities which are outside the scope of this unit of assessment.
The School has continued to focus its research on coherent groups based on School priorities and staff interests. The ability to map the groups neatly onto RAE units of assessments and criteria has not been a factor in deciding School policy or the groupings. Combined with the clear identification of staff who are research active, as opposed to scholarly active, has significantly reduced the number of staff submitted in this UoA compared to 1996. This reduction in the number of staff submitted does not indicate a reduction in the overall number of staff undertaking research in computing within the School. The number of active research staff has not changed significantly since 1996. Areas of School research not included cover animation, teaching and learning environments, special needs computing, interactive multimedia and individual research. As well as reducing the numbers of staff submitted, this continued focussing on priorities and needs, as opposed to RAE UoA and criteria, has also had an impact on the external income and the number of research students returned in this submission.
Changes in the School’s academic profile, including major developments in visualisation and virtual environments, have been reflected in the change in groupings of research staff within the School. In particular, the previous groupings of Modelling and Simulation and Visualisation and Interactive Systems have been replaced by Virtual Environments and Interactive Systems. This change has been helped by the impact of the JREI funding which encouraged interdisciplinary and multidisciplinary work between the Schools of Computing and Mathematics, Science and Technology, Social Sciences and Health. The success of the developments in the area of virtual environments has been recognised by the University as a major growth area and the University funded a Chair in this area in 1998. The University has committed to support this area in both funding for teaching and research.
The increase in interdisciplinary and multidisciplinary research has been encouraged by both the JREI and the University establishing a research fund specifically to support cross-School research activities. The following projects have all benefited from this support: -
§ computing and physiotherapy in the 3-D reconstruction of scoliotic body shape,
§ computing and psychology in the experimental content evaluation of virtual environments,
§ computing and virtual reality modelling of the built environment and
§ computing with social sciences to assess how virtual environments may be used to help people with dementia.
University and School semester sabbaticals (on average three per year) have been targeted at junior staff who have shown promise and wish to consolidate their research.
As part of the University’s revised policy on research, all members of staff wishing to supervise research students must attend the University’s Research Supervisors’ training course. This is part of a programme of events introduced to offer improved support for research students. Staff are always invited to provide a self assessment of their research activities via the personal development and review process that operates across the University.
In addition to the staff returned in this unit of assessment, the overall research culture and environment within the School is enhanced by groups working in areas outside this submission and colleagues undertaking individual research. For example, one member of staff (Dr McElhone) is working closely with York University on safety critical systems while another (Dr B Oates) is working with Salford University on information systems theory.
Research Groupings
The School’s research groupings are now centred on Formal Methods, Virtual Environments and Interactive Systems and Medical Informatics. Cross-references to RA2 outputs are given in the form [SED-1] indicating the first output for S E Dunne, for example.
Formal Methods – S E Dunne (SED) and W J Stoddart (WJS)
The Formal Methods group has interests in model-based formal specification languages, integrated formal methods, general correctness and object technology. Particular areas of work are
FM1 Extending B-GSL to allow concurrent components to be introduced during development (SED, WJS)
By adding predicate transformer rules for weakest liberal pre-conditions the group has extended the semantics of B-GSL from total to General Correctness semantics. The practical aim of this exercise is to simplify proof obligations associated with the introduction of concurrency during refinement. [SED-2]
FM2 Recasting Hoare and He's Predicative Program Model in the Context of General Correctness (SED)
By reformulating Hoare and He's Unifying Theory in terms of General Correctness we have been able to express their healthiness conditions for a program entirely algebraically in terms of zero and unit laws.
FM3 Integrated Formal Methods (WJS)
We have developed two approaches to combining state based and behavioural views of event based system models. These are ZCCS, a variant of CCS which has Z as its value passing calculus, and the "Event Calculus", a graphical notation based on state machines which communicate via shared events, and which can be translated into Z or B. The state machine approach allows desirable system properties, such as freedom from deadlock, to be expressed in a system invariant which gives rise to proof obligations in the style of the B method. [WJS-1, WJS-2, WJS-3]
FM4 Execution Architecture for an Abstract Command Language (WJS)
We have described an efficient virtual machine architecture which supports abstract programming concepts such as non-deterministic choice. It also supports the introduction of concurrency during refinement, as mentioned in FM1 above. As part of this work, we have reported a variant of the probabilistic choice defined by Caroll Morgan in his formalism pGSL. It differs from Morgan's probabilistic choice in complying with the predicate transformer rules of non-deterministic choice - this allows probabilistic choice to be combined with backtracking. We have shown that, like pGSL, our variant satisfies the important property of "sub-linearity", from which other properties such as monotonicity and scaling may be derived.
FM5 Case Studies (SED, WJS)
Several case studies have been published based on the above techniques, including a distributed seat booking system with timeouts, an invoicing system, a gas burner control system with timing constraints, and a hybrid system with environmental components that evolve in continuous time. [WJS-1, WJS-4]
FM6 Specification Languages: Z and B (SED, WJS)
In addition to our use of these notations in FM1 to FM5 we have investigated i) problems relating to undefined terms with reference to underlying denotational semantics, proof theory and historic approaches such as that of Russell in Principia Mathematica, and ii) different uses of the concept of a state-invariant. [SED-1, SED-3, SED-4]
Virtual Environments and Interactive Systems – Professor P G Barker (PGB), Professor M O Cavazza (MOC), P C Fencott (PCF), Dr W Tang (WT) and Dr A Wani (AW)
The Virtual Environments and Interactive Systems group has strong interdisciplinary and multidisciplinary links with researchers from other Schools within the University as well as other national and international groups. Professor Phil Barker leads the long established Interactive Systems component of the group and Professor Marc Cavazza leads the Virtual Environment component of the group. Professors Barker and Cavazza have extensive experience of national and European research projects.
Members of the group have been involved with external exchanges with overseas universities in America, Europe and Japan as well as being invited speakers at international conferences.

A significant part of the group’s research deals with the notion of Virtual Actors. This includes several aspects, from motion control [MOC-4] and behavioural animation [MOC-1] to conversational characters [MOC-3] or "talking heads". Most of this research is interdisciplinary and involves the combination of Computer Graphics, Virtual Reality, Artificial Intelligence, Usability, and Human Factors.
VI1 Language-enabled Agents
There is a strong relationship between the development of intelligent agents and their ability to communicate with the user using natural language. On one hand, the execution of complex tasks is best controlled using high-level concepts rather than direct control. On the other hand, while understanding natural language brings incomparable benefits in terms of user-friendliness, language-enabled agents should also be able to generate low-level behaviours [MOC-3].
VI1.1 Conversational Characters (MOC)
There is a growing interest in conversational characters as interface agents and intelligent assistants, both in traditional interfaces and 3-D environments. Most of our research in this area has taken place through an Interactive TV project.
The "Virtual Interactive Presenter" (VIP) is an EPSRC/DTI LINK broadcast project in Interactive Television involving Cambridge University Engineering Department (Speech Recognition), the BBC (on-line programme guide), Sony Network Solutions Europe (software architecture) and Advance Multimedia Communications plc (character modelling and animation, user interface, database). The goal of VIP is to develop a conversational character serving as an interface to an Electronic Programme Guide. The School is in charge of the overall project co-ordination as well as developing the Natural Language Processing (NLP) modules (Human-Computer Dialogue).
The project has resulted in the development of a fully implemented prototype, integrating speaker-independent speech recognition, human-computer dialogue and real-time animation of the character. The underlying dialogue technology is based on the identification of speech acts from the comparison of the semantic contents of successive user utterances. This approach has been presented in specialised workshops and conferences in Human-Computer Dialogue.
VI1.2 Natural Language Instructions (MOC)
This research investigates the use of natural language to control the high-level behaviour of embodied agents in virtual environments. Apart from addressing some fundamental issues in the integration of NLP and computer graphics, this research has applications in computer animation, computer games, simulation and interactive storytelling. This research has been conducted through the development of various prototypes and subsequent experimentation. The following results have been obtained [MOC-3]
§ the processing of instructions referring to spatial tasks requiring the ability to parse a sub-language including spatial expressions. We have implemented this feature through Tree-Adjoining Grammars.
§ the integration of NLP and graphics poses specific problems, for instance, in the case of reference resolution. A model of deferred reference resolution has been proposed. Also, when controlling an animation system, there is a need to generate low-level commands from the semantic content of the sentence. We have developed a technique using a pivot language (structured command language) interpreted by the animation system.
§ there are specific real-time issues concerning most interactive 3-D applications. We have optimised the NLP steps so that the parser can run in real-time (< 100 ms to parse a 15-word utterance, speech recognition excluded).
VI2 Agent Behaviour and Avatar Control
It is possible to classify artificial actors depending on their level of autonomy, from Avatars, which are direct embodiments of the user, to autonomous actors that exhibit intelligent unsupervised behaviour. Both user control and autonomy require specific techniques [MOC-2].
VI2.1 Path Planning (MOC)
Path planning for virtual actors in virtual environments is a key technique for autonomous spatial behaviour. There are several techniques for path planning, though recently attention has focussed on search-based path planning, where search algorithms are applied to discretised geometric environments. We have extended these techniques to
§ the co-ordination of small group of agents, through the notion of synchronised path-planning through non-admissible heuristic search
§ the use of semantic information in the search algorithm, in addition to simple distance (geometric) heuristics
VI2.2 Behaviour-based Animation (WT)
This work studies and develops techniques in the area of behavioural control and animation for interactive three dimensional computer virtual environments. Two major areas of investigation are artificial sensing techniques [WT-4] for acquiring information from the environment and the knowledge processing techniques for decision-making.
A simple, yet effective, synthetic vision technique for autonomous object’s vision in the virtual environment has been developed and two primary test bed systems that implement the vision technique have been developed.
Current work involves
§ the investigation and implementation of non-deterministic behaviour modelling techniques [WT-1]
§ the implementation of advanced motion control techniques to control an object’s motion in the virtual environment [WT-2, WT-3]
§ the investigation of advanced computer graphics techniques to simulate the changes of the virtual environment in real time so that the object’s behaviour would change accordingly
§ 3-D computer graphics modelling techniques incorporating real-time rendering techniques.
VI2.3 Motion Control of Avatars (WT, MOC)
Physical control of an avatar is the best known modality of user control and is currently used in animation, virtual reality and computer games. Physical control can be based on motion capture using various sensors. In this case, inverse kinematics can be used to minimise the number of sensors. However, there exist limitations in the real-time integration of motion equations, which limit the recognition of fast gestures. We have proposed an inverse kinematics techniques implementing a fast Lagrange-like integration method that can cope with faster gestures than traditional methods [MOC-4, WT-2, WT-3].
VI3 Intelligent Virtual Environments
Virtual environments are essentially spatial structures in which users navigate, and interact physically with their constituent objects. As such, they come with little semantics attached. Intelligent virtual environments investigate various forms of high-level behaviour and abstract representation in virtual environments.
VI3.1 Action Recognition (MOC, WT)
While emphasis in virtual environments is mostly on objects, there is little work dedicated to action representation and formalisms. The automatic recognition of actions is nevertheless required for scene interpretation and adaptive response to the user behaviour. Action recognition in virtual reality is similar to scene recognition in computer vision. It should however be facilitated by the direct availability of low-level events in the graphic database.
We have produced an action recognition technique base on the parsing of low-level events through cascaded Finite-State Transition Network (FSTN) [MOC-1]. The procedure by which actions to be recognised are described by FSTN is a form of under-specified action representation derived from a plan-like decomposition of an action. Several applications have been proposed in conjunction with geometrical path planning or in virtual theatre.
VI3.2 Virtual Environments and AI Techniques (AW)
Research has been undertaken in basic AI techniques and their use in various applications including developing synergistic and distributed synergistic neural network models. A new inductive learning algorithm called SAFARI has been developed.
We have applied intelligent techniques in optimal multi-resolution model building [AW-3] with further research plans to develop optimal intelligent virtual environments for virtual reality and computer entertainment applications.
Intelligent techniques have been applied in various pattern recognition tasks [AW-1, AW-2] with further research plans to apply hybrid intelligent techniques in pattern recognition tasks and intelligent virtual environments.
A new algorithm in the area of machine learning has been developed with further research plans to develop appropriate knowledge representation techniques to support multi-strategy task adaptive inferencing [AW-4]. The outcome of this research will be used to extend further intelligent virtual environments.
VI3.3 Virtual Environment Theory (PCF)
The virtual environment theory team is a multidisciplinary team of academics with specialisms in virtual environments and computer science, psychology and HCI, and aesthetics of interactive digital media. The team developed out of an earlier grouping, which undertook research into methods integration for software engineering [PCF-4]. The team is developing perceptual modelling techniques for virtual environment content [PCF-2] and conducts experiments to ascertain the correlation of user behaviour with the predictions of the models. This research allows for the comparative content analysis of a wide range of diverse virtual environments [PCF-3], which in turn is leading on to the development of design methods for virtual environments which focus on both the engineering and the aesthetic aspects of virtual environment design [PCF-1].
VI4 Interactive Systems
Our European collaborative work on the application of neural networks to pattern recognition resulted in the award of an MPhil (1996) while our work on Distributed Performance Support Systems concluded in 1998 with the award of a PhD (1998). Our research into Hypermedia Electronic Books [PGB-2] has also finished culminating in the award of a PhD (1998). Related to this was our investigation into various cognitive and usability issues of Hypermedia Authoring Tools. This work successfully concluded with the award of a PhD (1998).

VI4.1 Completed Projects (PGB)
Our work on digital libraries, online learning, icons and mental models continued throughout this period. Papers were presented at various international conferences and, perhaps, most importantly our new book on "Iconic Communication" was published in 2000.
During this period work has been undertaken on behalf of the Open University, Strathclyde University, University of the West of England, De Montfort University, the Institute of Learning and Teaching and SEDA (Staff and Educational Development Association).
VI4.2 International Projects (PGB)
We are currently working on four international projects. Three of these involve PhD students who are registered at the University of Teesside. An Australian Project involves an investigation of the creation of models and methodologies to support the development of interactive multimedia learning products. Our American project is studying the efficacy of different modes of electronic course delivery for distance learners who are resident in the state of Nevada. Our Lebanese project is exploring the role and potential of a virtual university system for the delivery of electronic courses within Lebanon. Our project with the University of Barcelona [PG-1, PG-2, PG-4] is exploring the issues involved in collaborative distance authoring of hypermedia materials for delivery to distance learners via the Internet.
Some of our current UK projects include work on European Patents (HCI issues), an evaluation of computer conferencing techniques, icon research and the development of multimedia patient education systems.
VI4.3 Future Work (PGB)
A book on "Web-Based Teaching and Learning" is planned, as is a book that looks at "Applications of Iconic Communication". New methods for the support of online learning and electronic course delivery, especially with respect to the incorporation of digital library facilities, continue to be explored. We are particularly interested in automatic portfolio management and assessment techniques.
Medical Informatics – Dr J J Longstaff (JJL) and Dr D S Simpson (DSS)
The group’s research activities are in close collaboration with clinicians in both the primary (General Practice) and the secondary (NHS Trusts) healthcare sector, healthcare authorities as well as medical physicists. The group has long established links with the NHS Information Authority (NHSIA) and its predecessors. One Visiting Professor and two Visiting Fellows from the local NHS Trusts are associated with the group.
MI1 Electronic Patient Record and the Electronic Health Record (JJL, DSS)
A major strand of the work over the last few years has been concerned with the Electronic Patient Record (EPR) and the Electronic Health Record (EHR). In particular, we have developed a model of confidentiality for the medical record. This model and our work on the requirements of a multimedia medical record has been the subject of lengthy discussion with various parties including the NHSIA and its predecessors. It was pleasing to see a number of the items worked on with the NHSIA appear in the Government’s Information for Health - An Information Strategy for the Modern NHS 1998–2005 when it was published in 1998.
The group is a partner of one of four winning bids to pilot examples of pan-community EHRs as part of the NHSIA’s Electronic Record Development and Implementation Programme (ERDIP) following the release of the Government’s Information for Health strategy. A key component of this project is based on the confidentiality model of the EHR developed by the group. The model is a development of earlier database research with which members of the group have been involved for many years.
Alongside the pan-community EHR project, work started in late 2000 on an externally funded project on direct booking in conjunction with a local NHS Trust. This multidisciplinary and interdisciplinary project involves research into the integration of data from both primary and secondary care systems across the NHSnet, the capturing of endoscopy data from the operating theatre via a spoken dialogue system and access controls to the data to ensure confidentiality. The first stage of this project, the work concerning the spoken dialogue system (see MI2 below) has started evaluation trials in a local NHS Trust. [JJL-1, JJL-2, JJL-3, JJL-4]
MI2 Spoken Dialogue Systems (DSS)
We have developed a spoken dialogue system for the hands free recording of endoscopic information. The language processor uses a lexicon and a domain-specific semantic grammar. The grammar rules are expressed in the style of a Definitive Clause Grammar which specify rules of semantic transformation as well as syntax. Domain item knowledge is structured into a slot-filler notation while domain anatomy knowledge is structured into a hierarchical model of anatomical features.
Early results have been very encouraging and this work has already been extended to demonstrate voice control of a surgical robot manufactured by Armstrong Healthcare Ltd. Discussions are being held with Armstrong Healthcare and clinical interests to extend this work further into a general spoken dialogue system for the operating theatre (e.g. lights, table, robots, etc.). [DSS-2, DSS-3]
MI3 Bone Mineral Densitometry Reference Ranges (DSS)
A major piece of research completed by the group since the last RAE has been the development of a proposed methodology for the construction of national bone densitometry reference ranges. The methodology developed was used to construct national bone densitometry reference ranges utilising a study of 1,372 Caucasian women involving data collected from leading hospitals in London, Manchester, Birmingham and Leeds. [DSS-1, DSS-4]
MI4 3-D reconstruction of Scoliotic Body Shape (DSS)
An exploratory piece of collaborative work started in 1998 with two junior members of staff (one in the School of Computing and Mathematics and one within the School of Health) and a consultant colleague to investigate the 3-D reconstruction of scoliotic body shape. This early work has shown sufficient promise to obtain funding (£60K) from HEFCE in 2000 for the purchase of a 3-D scanner to develop this work further.
MI5 Workshops and Meetings (JJL, DSS)
The group holds regular monthly meeting involving academics, clinicians and healthcare workers at which research topics of mutual interest are discussed. An annual Innovations in Healthcare Computing workshop is held which attracts between sixty and seventy delegates. A part-time MSc in Medical Informatics, developed by the group, is delivered jointly by the School and clinical colleagues.
Summary
The work described in this UoA is embedded within a much larger research base and culture within the School and benefits from strong support from the University’s most senior management. The focus of the School’s research has evolved since 1996 and has benefited significantly from the University and School review undertaken 1998/99. It is expected that the groupings described here will continue to develop at an increasing rate over the next few years with a potential for one or two new groupings forming over the next five years. Of the existing groups, the Formal Methods group and the Virtual Environments and Interactive Systems group both have a strong international dimension. The Medical Informatics group has a strong national and collaborative Government agency element, with its work presented at an international level.

Users of this website should note that the information is not intended to be a complete record of all research centres in the UK

Copyright 2002 - HEFCE, SHEFC, ELWa, DEL

Last updated 17 October 2003

[ Home | About the RAE2001 | Results | Submissions | Overview reports | Panels | Guidance for panel members
| Guidance for institutions | Publications  ]