Research Domain |
Academic Staff |
Associated Technologies |
Researchers |
Publications |
Current Projects |
|
Collaboration |
Links |
Faculty of Computing, Engineering and Mathematical Sciences
University of the West of England, Bristol
Advanced Computation in Design and Decision-making
The integration of Computational Intelligence and supporting computing / computational technologies with complex human and machine-based design and decision-making environments
|
Research Domain |
Academic Staff |
|
Associated Technologies |
Researchers |
|
Publications |
Current Projects |
|
Collaboration |
Links |
The objectives of the ACDDM group concern the generation of high-quality information relating to complex design and decision-making domains. Such domains are typically characterised by uncertainty and poor problem definition where high levels of complexity confound understanding and limited domain knowledge may initially restrict meaningful machine-based problem representation. Qualitative aspects of the problem area may demand a high-level of user-interaction in terms of complete or partial user-assessment of generated solutions. Frequent redefinition of any machine-based problem representation is necessary as information gained leads to a better understanding of complex problem characteristics.
A human-centric approach is taken where experiential and arising user knowledge is integrated with powerful machine-based search and exploration. Concurrent, adaptive, non-linear search processes are proposed that generate diverse high quality solutions plus a wealth of background information appertaining to problem characteristics and the nature of the complex problem space. A further challenge is the extraction and processing of such information and its succinct presentation to the designer / decision-maker. The hypothesis is that off-line assimilation of such information contributes directly to the decision-makers' knowledge base and supports the further development of machine-based representations. Knowledge capture is thus intimated through the integration of experiential and arising user knowledge through further, iterative search of an evolving problem space. Objectives associated with this process relate to the development of highly user-interactive systems that support discovery, innovation and creative activities.
Developing interactive systems support and enhance the inherent skills and capabilities of the designer / decision-maker rather than attempting to replace them through the development of entirely machine-based processes. Although the curse of dimensionality presents extreme difficulties to the human decision-maker it is proposed that such difficulties can be overcome through an appropriate melding of machine-based search, information generation with human intuition, judgement and experiential knowledge. An inherent objective is to develop machine-based processes that relieve the cognitive load upon the user through a reduction of problem dimensionality and the provision of high-quality information. Innovative and creative activity is supported through appropriate human / machine interaction.
Design is considered here to be a generic activity. The process of design involves and impacts upon a diverse set of problem-solving and decision-making areas. Although previous research has concentrated upon the various stages of engineering design (Parmee I. C., 2001) areas currently under investigation involve compound design and drug discovery, portfolio design; supply chain design and management, schedule and resource allocation and software design and engineering. Each of these problem areas exhibit similar characteristics in terms of uncertainty, poor initial problem definition, a requirement for both qualitative and quantitative judgement, non-linearity, high modality within multi-variate space plus heavy constraint and multiple criteria.
The contribution of Computational Intelligence to systems under development relates to Evolutionary Computing (EC) which provides stochastic search processes that can efficiently negotiate complex spaces described by mathematical, statistical, boolean, neural network or fuzzy inference models. Machine-learning techniques that are currently contributing to on-line learning requirements. Software Agent technologies are providing background processes relating to a number of tasks whereas multi-agent systems are providing a system modelling and machine-based negotiation capability. A major challenge relates to the achievement of autonomous agent activity through appropriate on-line machine-learning. Statistical, neural and fuzzy modelling techniques can also provide nimble approximations to supplement results from computationally expensive analytic methods. Cluster-oriented Genetic Algorithms are currently providing a Data-mining capability whilst other established techniques are under investigation. Various search space sampling algorithms are also contributing in this area.
Human-computer Interaction (HCI) is receiving a deal of attention to address human factors in relation to complex systems whilst supporting appropriate graphical user interface development and related human-centred issues. Data Visualisation is an essential component supported by agent-based data processing to ensure succinct information presentation to the user. Novel approaches both in terms of visualisation and data processing are under investigation and complement a range of established techniques. It is apparent that to develop anything approaching real-time interactive intelligent systems High-performance Computing capabilities will be essential and current involvement with the e-Science research community investigating distributed problem solving is contributing to research in this area. The ability to access and utilise major distributed computing capability is seen as essential.
The Research Group comprises the following full-time academic staff:
Professor Ian Parmee, Group Director; Research Leader, School of Computer Science
Professor Larry Bull, School of Computer Science.
Mr Barry Dean, Senior Lecturer, School of Computer Science
Mr Christopher Simons, Senior Lecturer, School of Computer Science
Dr. Praminda Caleb-Solly, Lecturer, School of Information Systems
Current
Azahar Machwe, AHRC Research Associate, School of Computer Science
Alistair Sedwell, Research Student, Evotec OAI / School of Computer Science
Chris Simons, Faculty Research Student, School of Computer Science
Barry Dean, Faculty Research Student, School of Computer Science
Previous
Dr. Mark Shackelford, DTI / EPSRC Research Fellow, School of Computer Science
Dr Johnson Abrahams, formerly Faculty Research Student / Associate, School of Computer Science
Bhuvan Sharma, Research Associate, School of Computer Science
Industrial / Commercial Collaboration
Members of the group have a history of collaboration in terms of research and / or consultancy with:
Alsthom
BAE Systems
British Energy
Evotec OAI
Lafarge Braas
LloydsTSB
Perkins Technology
PricewaterhouseCoopers
Rolls Royce plc
Rolls Royce Associates
SEA Ltd
North West Regional Development Agency
_________________________________________________________________________________
People-centred Computational Environments for Design Discovery: Establishing a Fully Integrated Framework
Funding body / scheme: AHRC / EPSRC Design for the 21st Century Initiative
Principal Investigator: I. C. Parmee, Bristol UWE.
Co-investigators: E. Hall, Cambridge University; J.C. Miles, Cardiff University; J. Noyes, Bristol University; C. Simons, Bristol UWE; D. Smith, Newport University
Researchers: A. Machwe, Bristol UWE; D. Williams. Bristol University.
The research concerns continuing support for the research activities of a highly multi-disciplinary community of academic and industrial practitioners who, since March 2005, have developed a strong working relationship within the EPSRC / AHRC ‘Discovery in Design (DiD: People-centred Computational Issues’ Design for the 21st Century Cluster which facilitated a cross-disciplinary understanding of possible basic requirements for a generic, people-centred computational environment to support early formative stages in design via an initial exploration of multi-disciplinary requirements, possible computational structure and people centred-issues relating to complex, generic conceptual design systems [Parmee et al., 2006].
We are continuing the research with a 12 month study, whose objectives are to:
i) establish a completely integrated, initial framework for people-centred computational environments that support conceptual design across multiple, diverse disciplines;
ii) develop a detailed definition of research requirements and agendas that will ensure short to medium term benefits through interim development whilst ultimately leading to the eventual realisation of generic commercial systems.
The study will continue to stimulate new ways of design thinking in terms of highly novel, user-interactive computational architectures. These embrace established and emerging computational intelligence technologies to provide the necessary support in a domain where initial uncertainty, poor problem definition and high risk are inherent.
In terms of international significance, those economies that invest in conceptual design support will ultimately succeed within an increasingly competitive global market through the identification of innovative, creative and highly competitive design solutions. An opportunity exists to gain a competitive edge through development of conceptual design support that allows early-stage designers access to massive computational capability currently unavailable due to lack of appropriate systems.
_________________________________________________________________________________
EPSRC Network:
People-centred Computational Environments for Design and Decision-making
Funding body / scheme: EPSRC Research Networks
Principal Investigator: I. C. Parmee, Bristol UWE.
Co-investigator: J. Noyes, Bristol University
The Network provides support for a highly multi-disciplinary community of academic and industrial researchers / practitioners. A group who have forged a strong working relationship within the 12 month ‘Discovery in Design: People-centred Computational Issues’ Cluster (EPSRC / AHRC ‘Design for the 21st Century’ Initiative) provide the initial core for this larger, growing community. Cluster membership was limited in order to ensure productive Workshop discussion across a highly complex, multi-disciplinary domain and the achievement of stated objectives over such a limited period of time. This strategy has proved to be highly successful. The aim of the Network is to now support a significant expansion by moving the focus of the group from design processes to the early stages of decision-making processes in general involving further disciplines. The result is a highly multi-disciplinary community with a far broader remit, a wider range of activities and a larger membership.
This is in line with the stated Cluster proposal objective i.e. ‘In terms of continuation an EPSRC Network proposal will be developed over the final few months of the cluster funding. The objectives of the Network will be to maintain the cluster activities and develop the research community growing from the areas identified’.
To ensure further continuation post-network funding, the Network has established the Institute for People-centred Computation (http://www.ip-cc.org.uk) which provides a focal point for collaborative proposals; joint publication and dissemination; workshops, summer schools, special interest groups, training events and close industrial / academic working relating to the development of people-centred computational environments to support early design and decision-making processes.
Publications
see http://www.ad-comtech.co.uk/Parmee-Publications.htm
________________________________________________________________________________
Interactive Evolutionary Design and Decision Making in Software Engineering Design.
Research Student: Chris Simons
Director of Studies: Ian C. Parmee
Objectives: To support and enhance software engineering analysis and design by the integration of Interactive Evolutionary Computing Design and Decision Making strategies and processes with multi-variate, complex and ill-structured software design domains.
Synopsis:
The research has identified commonality between the two design fields of engineering design and software engineering, taking Pahl and Beitz as an engineering design example and the Unified Software Development Process (Jacobson, 1999) as an example of software engineering. It is hypothesised that software engineering will benefit significantly from engineering design’s approach to solution variant generation and selection within conceptual and embodiment design to improve the overall "goodness of fit" and quality of software design. With this in mind, it is further hypothesised that sufficient commonality exists between engineering design and software engineering to enable the integration of interactive evolutionary engineering design strategies and processes with software engineering in a human-centric design environment.
Software engineering analysis and design is under investigation with particular attention to the representation, dimensionality, geometry and metrics of software designs, including the cognitive psychology of functional reasoning underpinning the analysis and design of software when derived from multiple requirements. It is intended that this research explores in detail the characteristics and semantics of "goodness of fit" of software solution design variants, making them amenable to their generation and selection against multiple conflicting objectives and constraints by evolutionary computational techniques.
Publications
Simons CL, Parmee IC, Coward D: 35 years on: to what extent has software engineering design achieved its goals? IEE Proceedings – Software, 150 (6), 2003, pp 337-350.
Simons CL, Parmee IC: Multi-objective Genetic Algorithms in Object-Oriented Conceptual Software Design. Procs. Int Conf. on Artificial Neural Networks in Engineering, St. Louis, US; 2006
Simons CL, Parmee IC: A cross disciplinary technology transfer of search-based evolutionary computing: from Engineering Design to Software Engineer Design. Journal of Engineering Optimisation; July 2007.
Others to follow - see http://www.ad-comtech.co.uk/Parmee-Publications.htm
____________________________________________________________________________________________________
Adaptive Search and Exploration Environments for In-silico Drug Design and Discovery
Research Student: A. Sedwell
Director of Studies: I. C. Parmee
Objectives: To develop chemically aware evolutionary / agent-based search and exploration techniques that can deal with the complex nature of chemical space utilising user-centric approaches to solution evaluation and the satisfaction of both quantitative and qualitative objectives. Overall development of experimental strategies utilising various computational intelligence technologies to support scientific decision-making in the drug design and discovery domain.
Synopsis: Initial research involved a comprehensive literature search and analysis of current techniques relating to the use of evolutionary computing for de novo molecule design [1,2]. This study reviewed different techniques employed to handle the problems associated when searching within chemical space and identified problems concerning current techniques.
Subsequently, appropriate evolutionary / stochastic de novo molecule generation algorithms and multi-agent based approachs have been designed to find practical solutions that satisfy multiple objectives. This work has included the examination of molecular representation and the selection and interfacing of appropriate fitness functions. Close communication with Evotec scientists and in-house testing of the developed approaches has contributed to the creation of suitable human interfaces to investigate partial solution evaluation by the user. Infrastructure development has involved appropriate interfaces to data libraries and appropriate distributed computing processes that can handle the computational expense deemed to be inherent in the integration of evolutionary search and exploration with drug design. Development of appropriate user interfaces, responses to user feedback and analysis of generated results will follow. A concurrent investigation of the contributory potential of e-Science technologies via close communication with the Cardiff / UWE DIPSO research project has also taken place.
Further investigation will focus on the development of dedicated user-centric approaches to the satisfaction of qualitative multiple objectives. This will aim to feedback useful knowledge to users and an important part of this work will be to conceive sophisticated user interfaces specifically for analysing the results from both agent-based and evolutionary molecular design.
Further study will concentrate upon human-computer interface aspects of the previous work involving the integration of developed experimental techniques and strategies with current de novo molecule design practice within Evotec and a detailed investigation of the manner in which subjective evaluation based upon scientific experiential knowledge can be seamlessly integrated with machine-based search and exploration.Publications
Others to follow - http://www.ad-comtech.co.uk/Parmee-Publications.htm
____________________________________________________________________________________________________
Agency-based Integration of Aesthetic Criteria with Evolutionary Design Search and Exploration
Research Student: A. Machwe
Director of Studies: I. C. Parmee
Synopsis
The background for this research is user-centric intelligent systems and creativity in design. Creativity is initially considered through the inclusion of aesthetics as additional design criteria within a semi-autonomous machine-based design environment. The system proposed brings together the fields of agent-based machine learning, stochastic search and subjective evaluation in design space search and exploration for aesthetically pleasing, structurally feasible and usable designs. It is intended that the system learns basic characteristics relating to aesthetic criteria from user-interaction with an evolutionary search process. This will result in a gradual lessening of the degree of user interaction allied with an increasing degree of autonomous machine-based solution evaluation. Thus such a system would during the initial stages be dependent on the user for guidance and direction and as the user crystallizes the concept in his mind and interacts with the system, it moves from user dependence to user assistance.
Although research relating to artificial design environments is evident in the literature there is no evidence of the integration of user evaluation, evolutionary search and exploration and agent-based machine learning With respect to the addition of aesthetics into computer-based design, much theoretical work (in the form of computer models) has been done in this field but very little application-based work is in evidence. We perceive three main components of the system namely the User, the Agents and the Evolutionary Search and Exploration System.
The User is the designer with little or no idea of the nature of the complex multi-variate and multi-objective design space. The primary purpose of the User is to provide the initial design requirements to the system and to evaluate the designs returned from it during the initial stages. The Agents have multiple tasks which include creation of the initial population based on design requirements (such as restrictions on placement of supports) and monitoring the designs for feasibility during the optimisation process. The machine-based system searches the space of possible designs identifying high–performance solutions based on multiple criteria such as structural stability and aesthetics and returns best solutions in terms of quantitative criteria for user-evaluation in terms of aesthetic criteria.
Publications
Machwe AT, Parmee IC: Multi-objective analysis of a component based representation within an interactive evolutionary design system. Journal of Engineering Optimisation; July 2007.
Machwe AT, Parmee IC: Integrating Aesthetic Criteria with Evolutionary Processes in Complex, Free-form Design - an Initial Investigation. IEEE Congress on Computational Intelligence, Vancouver, Canada, 2006 - BEST STUDENT PAPER AWARD ( CEC Stream) -
Machwe AT, Parmee IC: Introducing Machine-learning within an Interactive Evolutionary Design Environment. Proceeedings of Design 2006, Dubrovnic, Croatia.
Others to follow - http://www.ad-comtech.co.uk/Parmee-Publications.htm
____________________________________________________________________________________________________
Recently Completed Projects:
____________________________________________________________________________________________________
Multi-objective Information Generation and Presentation within Interactive Evolutionary Design and Decision Making Systems.
Director of Studies: Ian C. Parmee
Research Student: Johnson Abraham
Objectives: From a human centric evolutionary standpoint:- To support and enable a better understanding of complex relationships between variable and objective space. Such an understanding will relate to design and decision making environments involving multiple quantitative and qualitative objectives.
Synopsis:
The research concentrates upon interactive evolutionary design environments relating to the identification of high-performance regions relating to multiple objectives via both cluster-oriented genetic algorithms (COGAs) and co-evolutionary multi-objective genetic algorithms (CoMOGAs) . It is intended to study the characteristics and nature of high-quality information generated from these two techniques and investigate how such information can be best processed and presented to the user. The information will relate to both high-dimensional variable and objective space. The aim is to process and present this information in a manner that supports the user in the development of an intuitive understanding of the relationship between variable and objective space thus enabling the designer to rapidly assess the degree of difficulty likely to be encountered when attempting to satisfy multiple qualitative and quantitative objectives The initial area of investigation relates to the preliminary design of military aircraft. air frames but the study will progress to a number of conceptual design applications.
The study also involves appropriate data-mining and visualisation techniques and the development of novel information presentation. Single function agent support will play a significant role in the handling of multi-dimensional information processing and in the reduction of cognitive load on the user. A supporting role for multi-agent negotiating systems is also under investigation.
Publications
Parmee IC, Abraham JA: Supporting Implicit Learning via the Visualisation of COGA Multi-objective Data. Proceedings of IEEE International Congress on Evolutionary Computation, Portland, USA, 2004; pp 395-402 - BEST OVERALL PAPER AWARD -
Others to follow - http://www.ad-comtech.co.uk/Parmee-Publications.htm
__________________________________________________________________________________________________
Wide Area Distributed Problem Solving
Funding body / scheme: Dti / EPSRC core e-Science
Research Fellow: Dr. M. Shackelford
Researcher: Johnson Abraham
Director of Studies: Ian C. Parmee
Objectives
The establishment of requirements for intercommunications between a team of problem solvers. The interaction is supported through a Problem Solving Environment (PSE), which enables coupling of various computational and data components from different sites.
Synopsis
This two year project brings together the resources of two groups of researchers with extensive complementary expertise relating to design, decision support and general problem solving; computational intelligence and distributed computing, and companies which deal with complex multi-variate problems requiring distributed problem solving. This collaboration seeks to harness the benefits of a number of methodologies for modelling and solving such problems through the integration of services at geographically distributed centres. The integration is supported through Grid technologies and software tools, such as Globus, and a Problem Solving Environment to enable annotation and sharing of results. The project utilises the service abstraction to connect different stand-alone software together, and enables the dynamic selection of services on-demand. The project will involve Grid-enabling existing software at UWE Bristol (the Optimiser), and integration between this and a software providing problem simulation (the Modeller), achieved through a Globus-enabled Problem Solving Environment from Cardiff. The project will then use this environment to create a demonstrator based on problems from the industrial / commercial collaborating parties, which include LloydsTSB, SEA Ltd, EVOTEC OAI, Avantis, and Emorphia. It will also be used as a proof-of-concept to demonstrate the effectiveness of using the Grid as infrastructure for supporting conceptual and collaborative design – and used to demonstrate this technology to other interested participants. The emphasis is on generic multivariate problem solving where initial poor definition and uncertainty are inherent features. For example, problems relating to those decision-making processes encountered during the initial phases of conceptual design of products be they engineering, financial or otherwise. A particular aspect of the study relates to the manner in which data management infrastructure can be used to support this process
Publications
Others to follow - http://www.ad-comtech.co.uk/Parmee-Publications.htm
__________________________________________________________________________________
Evolutionary Computing in Drug Design
Research Associate: B. Sharma
Director of Studies: I. C. Parmee
Objectives: To follow
Synopsis: To follow
Publications:
see http://www.ad-comtech.co.uk/Parmee-Publications.htm
__________________________________________________________________________________
Other Associated Publications
Bull L., Wyatt D., Parmee I. C., 2002, Towards the use of XCS in Interactive Evolutionary Design. In: Poster Proceedings of the Genetic and Evolutionary Computing Conference, New York.
Caleb-Solly P., Smith J., 2002, Adaptive Image Segmentation Based on Visual Interactive Feedback Learning. Proceedings of Adaptive Computing in Design and Manufacture V; Springer Verlag.
Parmee I. C. and Abraham J. A., 2004, Supporting Implicit Learning via the Visualisation of COGA Multi-objective Data. Proceedings of IEEE International Congress on Evolutionary Computation, Portland, USA, 2004; pp 395-402.
Parmee I. C., 2002, Improving Problem Definition through Interactive Evolutionary Computation, Journal of Artificial Intelligence in Engineering Design, Analysis and Manufacture - Special Issue: Human-computer Interaction in Engineering, 16(3)
Cvetkovic D., Parmee I. C., 2002, Agent-based Support within an Interactive Evolutionary Design System. Artificial Intelligence for Engineering Design, Analysis and Manufacturing Journal; Cambridge University Press, Vol.16 No.5.
Parmee I.C., Bonham C.R., 1999, Towards the Support of Innovative Conceptual Design Through Interactive Designer/Evolutionary Computing Strategies. Artificial Intelligence for Engineering Design, Analysis and Manufacturing Journal; Cambridge University Press, 14, pp 3-16.
Parmee I.C., Cvetkovic C., Watson A.H., Bonham C.R., 2000, Multi-objective Satisfaction within an Interactive Evolutionary Design Environment. Evolutionary Computation. 8 (2), pp 197-222.
Parmee I.C., Cvetkovic C.A.H., Bonham C.R., Packham I., 2001, Introducing Prototype Interactive Evolutionary Systems for Ill-defined Design Environments. Journal of Advances in Engineering Software, 32 (6), Elsevier, pp 429 – 441.
Parmee I. C. (2001) Evolutionary and Adaptive Computing in Engineering Design. Springer Verlag, London.
Parmee I. C. (2002) Supporting Innovation and Creativity through Interactive Evolutionary Systems. Poster Proceedings Creativity and Cognition 4 Conference, University of Loughborough, CHI Conference Publications.
GECCO Workshop Series on Interactive Evolutionary Computation - call for 2003: http://www.ad-comtech.co.uk/Workshops.htm
Publication List - I. C. Parmee: http://www.ad-comtech.co.uk/Parmee-Publications.htm
Discovery in Design: People-centred Computational Issues. An EPSRC 'Designing for the 21st Century' Cluster http://www.ip-cc.org.uk/did