Contact us Heritage collections Image license terms
HOME ACL ACD C&A INF SE ENG Alvey Transputers Literature
Further reading □ OverviewContents1. Background and introduction2. Executive summary3. The case for the Programme4. Technical content and targets5. Cost and funding6. Management of the Programme7. Human resources8. Summary of RecommendationsAcknowledgements
CCD CISD Harwell Archives Contact us Heritage archives Image license terms

Search

   
InformaticsLiteratureReportsAlvey
InformaticsLiteratureReportsAlvey
ACL ACD C&A INF CCD CISD Archives
Further reading

Overview
Contents
1. Background and introduction
2. Executive summary
3. The case for the Programme
4. Technical content and targets
5. Cost and funding
6. Management of the Programme
7. Human resources
8. Summary of Recommendations
Acknowledgements

4. Technical content and targets

INTRODUCTION

4.1.1 Information technology helps man handle and use information, and the system designer's aim is to produce a machine that matches, complements and extends man's capability. The designer's problems therefore centre on how the human user interfaces with the machine, how the machine itself organises and processes the information fed to it, and how the machine can be built. These are complex and demanding questions for the machines of today and future IT products can be expected to be more complex. Most of the proposed programme is to do with managing this complexity, and exploiting the growing capability of modern computing systems to handle in real time interactions of great complexity.

4.1.2 Four main enabling technologies are seen as essential to progress in this field.

  1. Software engineering, the efficient specification and generation of the instructions for the machines, is crucial to success. The UK is good at software but must develop better tools if it is to stay competitive.
  2. The interface between man and machine is also crucial. Commercial success will come to those who make their complex (or simple) products truly acceptable to their users. Visual, speech and touch input/output devices are the interface instruments and these need further development. This development must be supported by a better understanding of the nature of communication across the interface.
  3. A prime aspect of the extension of this interface into the machine is the study of intelligent knowledge based systems (IKBS). We want more powerful information processing systems with a more effective transfer of human intelligence and knowledge to the computer. We also want computer systems that are easier to build and use, implying a better fit between humans and computers.
  4. The hardware to make these systems will be heavily dependent on advances in very large scale integration (VLSI) of circuits on silicon. VLSI gives the capability of efficiently interconnecting very large numbers of logic elements and these will be necessary for many applications, typical of which are new, powerful processors for IKBS or signal processing units for speech or picture recognition.

4.1.3 For clarity of presentation this Section breaks the field down into these four technologies as though they were separate entities. In fact the links between them are very close and progress on one is usually dependent on progress in one or more of the others. Essential to the whole programme is a high degree of co-operation. To facilitate this we propose the early establishment of a communication network connecting involved organisations.

SOFTWARE ENGINEERING

4.2.1 A strong domestic capability in the technology of building IT systems is the key to a leading position in the world market. Software is a fundamental component of IT systems and accounts for an increasing proportion of their cost. Correspondingly, the capability to design and build software in the most reliable and cost-effective way is a crucial element in establishing an important position in the world IT systems market.

4.2.2 The technology of reliable and cost-effective software generation is the discipline, still nascent, of Software Engineering. Today it is common practice to construct even large and application-specific systems using ad hoc techniques. The practicality of doing this is changing rapidly, as systems become more complex and the market becomes more competitive; the ad hoc approach is already showing signs of strain. Methods must change. Efficient production, an engineered approach to reliability, conformity with requirements, and economical development and operation must become the UK norm. The strategic objective proposed in this area is that the UK should become a world leader in this Software Engineering technology by the end of the 80s.

The Focus: Information Systems Factories

4.2.3 The programme proposed to meet this objective is focussed on developing Information Systems Factories (ISFs). The aim is to establish by the end of the decade a lead in the UK's ability to provide ISFs. This will imply their widespread use and, as appropriate, export. An ISF will be a computer system, both hardware and software, which provides an integrated set of tools for producing IT systems using software engineering techniques. It will be developed from successive generations of Integrated Programming Support Environments (IPSEs), starting with a consolidated set of the tools available today. IT systems produced through the ISFs will meet the parallel requirements of efficient production and operation, and of improved reliability and performance, which the competitive market of the 1990s will impose.

4.2.4 An ISF will contain major subsystems for:

The development and use of each of these components will represent a significant advance, in its own right, in IT systems production. Each merits a sub-programme for development with its own goals and products.

4.2.5 On a broader scale, the programme - while focusing on ISFs - must have a number of subsidiary objectives. Some of these may be achieved in the earlier stages of work. They include:

The Strategy: Exploitation, Integration and Innovation

4.2.6 The implementation of this programme requires industry to change from its current ad hoc approach to a fully-engineered and far more capital-intensive style of production. It calls for an acceleration of research activities and of the evaluation and real-life application of research results. It also needs to accommodate the changes in technical direction which experimental activities may indicate - but which cannot always be predicted - within the framework of the evolving ISF and of its industrial exploitation. It is important to define the key elements of the strategy under which the programme should be implemented in order to meet its objectives with the greatest benefit to technological change and to industrial practice. There are three complementary elements: exploitation, integration and innovation.

Exploitation

4.2.7 The purpose of the exploitation element of the strategy is to see that the software engineering technology available at each stage of the programme is effectively taken up and used by the UK IT industry. In the short term there is an urgent need to promote the more widespread use of the simple tools available today. The scope for immediate improvements in software quality and productivity is large, and the ground would be laid for effective use of the more complex tools of subsequent generations of IPSEs. Three parallel initiatives are needed:

Organisations thus supported would be required, in return, in using the tools to participate in evaluation and measurement activities of the programme.

4.2.8 In the longer term continued effort will be needed to ensure that the advances resulting from research are developed and exploited by industry. The long-standing gap between research and production must be addressed directly. It is proposed that, as new techniques are produced by research projects and incorporated into new generations of IPSEs, they should be released for general use. For this purpose a public Software Production Centre would be set up to incorporate and exploit the latest technology. From the first year of the programme, it would provide services and equipment to support:

4.2.9 The programme should also attack the problem of quantification of the improvements in productivity and software quality that can be gained through the use of software engineering. Effort must be devoted to designing measurement methods and applying them in practice. Emphasis on software quality will lead to the growth of software certification. It is proposed to establish a National Quality Control Centre for the purpose of developing certification techniques and applying them to software products. The National Quality Control Centre would be introduced slightly later than the Software Production Centre as the area of quality, although extremely important, is less well understood at present. The Centre would provide facilities for:

4.2.10 The activities of these two Centres will stimulate the market for software components and provide for certification of components to national standards. These products would be made available through a Software Products Brokerage Scheme, to be run by industry as a commercial venture after financial support from the programme in the early years.

Integration

4.2.11 The second element of the strategy, integration, provides the framework for the evolution of a full Integrated Programming Support Environment (IPSE). An IPSE will contain a compatible set of tools based on a methodology for all phases of system development and operation, supporting both technical and management activities. Advanced IPSEs will also support multi-language developments allowing the system designer a choice of languages for programming and the ability to introduce re-usable software components. Similarly, development in both hardware and software will be supported so that either may be chosen as the preferred implementation medium for particular system components. By comparison, tools that exist today are specific to particular stages of the system life cycle and are generally incompatible with each other. The even more demanding requirements for multi-language operation and joint hardware and software development lie far ahead. A careful strategy is needed in order to build successive generations of IPSEs leading to the ISF.

4.2.12 In the short term, it is recommended that UNIX be used as the basis for IPSE development: it requires enhanced database and communications facilities in order to fulfil this role, but its widespread use in the industry qualifies it as a suitable starting point. However, alternative developments should also be supported; UNIX should not be introduced as a mandatory standard.

4.2.13 In the longer term, development of the second-generation IPSE and ultimately the ISF will require substantial research and development in the areas of:

IPSE development must be planned not only to incorporate existing tools and techniques, but also to absorb the innovations produced by research.

4.2.14 Of interest are the techniques expected from the UK Ada/APSE (Ada Programming Support Environment) programme. Ada/APSE has not been costed as part of this programme since it is expected to start prior to this programme with its own management and funding arrangements. However, close technical liaison and - where appropriate - interchange between the two programmes should be encouraged.

Innovation

4.2.15 The third element of the strategy, innovation, both complements and underpins the exploitation and integration elements. The innovation strategy aims to ensure that appropriate research for the programme is commissioned and that its evaluation and adoption by industry are properly followed through.

4.2.16 Specific areas for long-term research have already been identified: quantification of software quality and productivity, development of formal specification and prototyping methods, verification techniques, automatic software generation and methods of re-using system components. Important contributions should also be expected from research into very high level languages, IKBS techniques and other means of dealing with increasingly complex systems. Alternative approaches in these areas should be positively encouraged; in research, it is dangerous to select one apparently promising avenue from the outset to the exclusion of others.

4.2.17 Adequate testing and exploitation facilities for research results are vital. It is recommended that the programme should commission applications of new techniques which should be brought into development use via the Software Production Centre. Overall, the scale of UK research must be increased to compete with that of other countries and the extent of collaborative research, particularly between universities and industrial organisations, must be encouraged to grow.

4.2.18 Recommended activities in the software engineering programme

For immediate exploitation
  1. 1. Commission database and communications enhancements to UNIX to serve as the basis for the 1st Gen. IPSE.
  2. 2. Commission specifications for implementation of a compatible set of tools for the 1st Gen. IPSE, with associated training material.
  3. 3. Support training and installation of the 1st Gen. IPSE in the Software Production Centre and in a number of other organisations.
  4. 4. Support alternative implementations to UNIX of the tools and/or training programmes.
To be initiated early for medium-term benefit
  1. 5. Establish a programme aimed at the quantification of software quality and productivity.
  2. 6. Create arrangements for collaboration and information interchange with Ada/APSE developments.
  3. 7. Establish a programme of R&D for the 2nd Gen. IPSE covering:
    • methodology
    • tools
    • database
    • hardware and software framework.
  4. 8. Support research in a number of areas, including:
    • very high level languages
    • language theory
    • development of re-usable components
    • exploitation of concurrency
    • links with IKBS.
For later initiation
  1. 9. Establish the National Quality Control Centre.
  2. 10. Inaugurate the Software Products Brokerage Scheme.
  3. 11. Purchase equipment for 2nd Gen. IPSE.
  4. 12. Install 2nd Gen. IPSEs in the Software Production Centre and other organisations.
  5. 13. Begin programme of integration leading to ISF (3rd Gen. IPSE).
Table 1 Software engineering programme milestones: Open facilities
Year
1 establishment of Software Production Centre (SPC)
2 SPC in full scale operation
definition and planning of Quality Control Centre (QCC)
3 Software Products Brokerage in commercial operation
establishment of QCC
4 SPC installs 2nd Gen IPSE
QCC in full scale operation
5 demand for QCC certified software products
6-10 SPC installs and operates 3rd Gen IPSE
automatic verification techniques adopted by QCC
Table 2 Exploitation
Year
1 portable set of compatible tools developed and promised
2 supported growth in use of software tools
3 software tools in wide use in industry
4 2nd Gen IPSE comes into use
5 2nd Gen IPSE in wide use in the community
6-10 ISF (3rd Gen IPSE) available and in wide use
Table 3 R&D
Year
1 current R&D directions strengthened, with industrial participation
2 well-established collaborative programme for technology transfer. applying new techniques to life-size projects
3 construction of 2nd Gen IPSE
4 R&D for ISF (3rd Gen IPSE) under way
5 early employment of IKBS techniques in 3rd Gen IPSE development
6-10 growth in semi-automatic program generation
trend towards modular construction of systems from software and hardware components

MAN/MACHINE INTERFACE

4.3.1 The interaction of the man with the machine is fundamental to information technology, nowhere more so than with IKBS. The subject has not been wholly neglected in the UK but lacks a coherent thrust and there is a near absence of work on input/output devices, a field of considerable commercial significance, with opportunities to be exploited on a selected basis. A substantial programme is therefore recommended both on the man-related problems and the machine-related problems. The man-related problems of most significance are those concerned with the nature of communication across the interface. Commercial success will come to those who are making sophisticated products truly acceptable to their users. The suggested programme falls under three headings: human factors; input/output devices (transducers); and speech and image processing.

4.3.2 Human Factors

  1. Human/System Communication
    1. Design of dialogue and explanation capabilities of systems: identify the needs, conventions and skills in successful human communication, e.g. dialogue structures and control strategies, redundancy, ability to switch contexts, etc. Apply these findings to the design of system dialogues.
    2. Input/Output modes: investigate which modes (e.g. text, speech, images) are most effective for users in different applications, and how the human factors problems which they present can be overcome in the design of systems.
    3. Communicative skills: identify the factors which make human experts effective at communicating their expertise to other people. Use this knowledge to design systems which operate as effective consultants/tutors.
  2. Human/System cognitive compatibility:
    1. investigation of current systems - (a) identify the cognitive mismatch between the user and the system. What is the user's understanding of the system's reasoning and knowledge structure? How confident is he in the system's advice/solutions? Can and does he intervene in the reasoning process e.g. by questioning the reasoning or volunteering new information, etc? (b) explore how well the user and system work together on a particular task and how this symbiosis can be improved.
    2. Analyse human problem solving behaviour in complex tasks. Use this information (together with that from (i)) to create systems which are compatible with human reasoning.
    3. Investigate the extent to which systems need to emulate certain aspects of human problem solving behaviour or where alternative techniques are more suitable.
  3. Human-to-system: Expertise transfer
    1. Develop a set of reliable tools/techniques/guidelines for rapidly and effectively analysing the decisions of human experts in order to produce a detailed, formal and accurate description of the decision process.
    2. Explore whether individual methodologies can be developed for different types of problem solving task e.g. diagnosis, classification, risk analysis, forecasting, etc.

The research of human factors described above overlaps some aspects of the proposals presented in our IKBS programme. They are also presented here under the heading of MMI to stress the human dimension of the required research.

Input/Output Devices

4.3.3 With the spread of IT the means of access are becoming increasingly important. Once the level of use by an individual exceeds a certain threshold, it becomes essential for that user to have his own terminal or workstation. Paper-based and electronic information systems will need to co-exist. This will require input/output devices like high resolution scanners and hard copy devices to convert between one representation and the other and improved displays with matching characteristics, on a one-per-desk basis.

4.3.4 An increasing proportion of the hardware costs in future systems will be in the terminals and associated input/output devices. This represents a large market opportunity for companies. To exploit it they will have to develop the technology to build high quality, low-cost flat displays, printers and scanners. The development costs will be high, particularly in the case of displays, where extremely advanced technology is required. The development of a multi-functional flat panel display should therefore be supported by the programme.

Speech and Image Processing

4.3.5 This area involves knowledge extraction and presentation to provide input and output for the knowledge manipulation processes within the machine. The aim is communication between untrained, non-expert users and the machine via methods that are natural for humans. The first step is to analyse the input signal to remove redundancy and identify a small number of parameters. Combinations of these parameters are then chosen as features to be used in subsequent stages of processing. Next, the extracted features are compared with stored references. A further stage of processing can then take place using, for example, syntactic techniques. The final component in the understanding system is a simple knowledge based system which uses the knowledge extracted by the previous stages. There may be considerable interaction and feedback between the various stages of the system. Signal synthesis involves basically the reverse process.

4.3.6 The preliminary feature extraction of the input signal (or feature modelling and synthesis for output) must be as accurate as possible, to provide a reliable man-machine interface. One aim of feature extraction is to reduce the data rate of the incoming signal as far as possible, but all the essential knowledge must be preserved. Only then can the features reliably be used in subsequent stages of processing. Another important property of the man-machine interface is that the flow of information has to be processed in real time.

4.3.7 For speech processing, the bandwidth is low enough for conventional VLSI technology to be able to solve the real-time hardware problems (signal processing chips suitable for speech analysis are already on the market). For image processing the bandwidth is many orders of magnitude higher. Solutions can only be found with highly parallel architectures, realised either with specialised modules or with general purpose programmable structures. In conjunction with the hardware work, it is necessary to develop programming languages able to reflect and exploit the parallelism of the general purpose architectures. Analysis techniques for speech signals are well understood. The speech processing programme should, therefore, cover:

4.3.8 There are many algorithms dealing with particular aspects of image analysis, feature extraction' and synthesis, but a lot of work is needed on algorithm engineering to determine the best combination for each application. Other work should include:

Table 4 MMI programme Milestones
Year
Flat Panel Display
3 100 pixel per inch; 2 tone
4 100 pixel per inch; writable; 2 tone
5 240 pixel per inch; writable; colour
Speech Processing
2 speaker verification
3 speech synthesis from text; large vocabulary connected word recognition
5 multi-speaker and user-adaptive word recognition; studies of real time and continuous speech recognition; controlled intonation and customised voices.
Image Processing
2 signature verification
3 image enhancement and compression; picture editing
5 document and cursive script analysis; study of image recognition and understanding; document and picture analysis.

IKBS

4.4.1 An intelligent knowledge based system is a system which uses inference to apply knowledge to perform a task. Such systems can be envisaged handling knowledge in many areas of human thought and activity from medical diagnosis to complex engineering design, from oil technology to agriculture, from military strategy to citizen's advice. The development of such knowledge based systems is widely regarded as the best means of expanding the application of IT to activities which today's computing technologies cannot approach.

4.4.2 An IKBS consists of a user interface connected by an inference engine to a knowledge base, the whole structure supported by an appropriate virtual machine. Such systems can vary in power according to the scale and variety of the knowledge they contain, the procedures which they make available and the tasks which they are able to perform. For many tasks a large quantity of knowledge is required and a powerful problem solving capability is needed to apply that knowledge to these tasks. Even a system to carry out a single task will, in a typical IKBS, require sufficient complexity to be able to work with large, incomplete and rapidly changing knowledge stores, and with tentative or heterogeneous inference procedures for exploiting this knowledge in reacting to varied and unreliable inputs in a changing environment.

The present situation

4.4.3 Based upon the achievements in artificial intelligence and in computer science, results have already been obtained in the development of simple but useful knowledge based systems. Sufficient in fact has been achieved to produce a consensus that more capable and useful systems can and should be developed. There are some fundamental and challenging problems to be tackled. In addition to experimental work to test concepts and validate hypotheses concerning the possible approaches to efficient solutions, the underlying theory needs to be further developed.

4.4.4 The research required is not only difficult, it is essentially long term. But valuable experience can be obtained now by developing and experimenting with systems that tackle immediate needs. Progress in IKBS requires the full and effective collaboration of research workers in academia and in industry and the bringing together of a number of disciplines. In addition to work in artificial intelligence, there is much to be done in cognitive science, in logic and semantics, and in advanced computer science relating both to hardware and to software. These combined disciplines must be applied to the particular specialised domains of many varied potential users, each with their own extensive domain knowledge.

4.4.5 Research is required on the means for formal knowledge representation and inferential problem solving which will include the development of functional languages, functional machines and smart data bases. Research is required into basic techniques of knowledge acquisition, representation and manipulation (for example in classification and concept formation or planning and modelling) and in related matters such as image perception, message interpretation and robot management.

4.4.6 Many of the requirements of efficient knowledge based systems overlap with other advanced IT developments. Overcoming the present limitations of the human interface offered by conventional technology will be essential. The way forward appears to lie in the use of natural languages and machine processes more akin to human thought processes. There is a clear overlap between this requirement and other aspects of improving the interface between men and machines.

4.4.7 The efficient use of a large knowledge base requires a good deal of inference processing capability. The limitations of single processor sequential machines are a major inhibition. New hardware and software architectures are necessary. The development of these new architectures should enable profitable exploitation of the present capability to manufacture devices with enormous potential for concurrent processing. Work on the development of such novel architectures and of the conceptual means to employ them is a matter of great importance and urgency. In this area the needs of IKBS development overlap the needs of other IT developments and closely relate to the advanced research in the VLSI programme.

The IKBS Programme

4.4.8 We propose a ten-year programme of research and development of IKBS, the aims of which are:

  1. To promote research in all aspects of IKBS
  2. To ensure development from the research results
  3. To stimulate production of development prototypes.

The strategic objectives of this programme are twofold:

  1. To enable UK industry, Government and other organisations to employ IKBS productively in their own operations in order to enhance their efficiency and competitive position.
  2. To enable the UK IT industry profitably to supply IKBS and other products incorporating IKBS to the market place.

Thus the programme must take account of the needs of IKBS users as well as of the producers and researchers of such technology. Indeed the involvement of users in the development of systems is essential. We propose a programme which will bring together research in academia and in industry.

4.4.9 Until now almost all of the research in these areas has been done by the academic community. Only recently has UK industry begun to take an interest. There are still only a few companies with a capability in this area. If the programme is to be successful there must be a substantial increase in the number of people trained in IKBS research, development, and production. The facilities for training such people must be enlarged so that an expansion in both the academic and industrial research communities can be achieved, and to ensure that there are enough trained and knowledgeable people in user organisations to apply the results.

4.4.10 A ten year programme is needed because the research required is difficult, and because technological innovation and transfer in this challenging area is inevitably time-consuming. Time is also required to build up the research and development community to the size required to derive the full benefits of IKBS for the UK. The programme is in three modules:

  1. build up of the infrastructure, and enlargement and organisation of the R&D community;
  2. support of programmes of research into all relevant aspects of IKBS;
  3. construction of a number of demonstrator systems applying research findings to particular applications through collaborative projects between industrial and academic researchers and demonstrator users.

Build Up of the Infrastructure and the Research Community

4.4.11 The qualified UK IKBS research community, though distinguished, is currently very small and its members are scattered. In order:

These IKBS support facilities must be co-ordinated with the proposed software engineering programme.

Research Programmes

4.4.12 Further work is needed to draw up a detailed programme of research. The areas for research can, however, be identified as follows:

  1. Research into the basic capabilities needed for IKBS which are in two groups:
    1. The internal capabilities of the IKBS for basic knowledge representation and inferential problem solving, including:
      • classification, concept formation
      • summarising, abstracting
      • selection, retrieval filtering
      • reasoning, use of heuristics
      • planning, modelling
      • learning, memorising
    2. Capabilities for interacting with the world external to the IKBS, including:
      • language analysis and production
      • image perception and generation
      • physical object sensing and moving
  2. Research into the construction of IKBS systems for executing tasks using generic knowledge representation and problem solving along with the basic capabilities. These tasks include:
    • scientific diagnosis
    • database questioning
    • teaching basic skills
    • content-determined document retrieval
    • office scheduling
    • robot management
    • picture matching
    • message interpretation
    • law consultation
    • program drafting
    • technical manual production
    • industrial process control
    • warehouse packing

    An important part of this research will be concerned with the means of eliciting knowledge from humans for incorporation into the IKBS systems.

  3. Research into the technologies, and, more broadly, into the architecture of the virtual machines required, inter alia, by IKBS. This includes work on such things as:
    • functional languages and functional machines
    • logical languages and logical machines
    • data flow machines
    • reduction machines
    • rule based languages
    • smart databases

    and, more generally, work on non-conventional languages, parallel processing, and all forms of data-driven and non-deterministic systems.

The IKBS programme will support research programmes in these areas carried out in academia, in industry, and in both jointly.

Demonstrator Projects

4.4.13 The choice and detailed specification of these projects cannot be made now for it must depend both upon the detailed research programmes and their anticipated potential and upon discussions with prospective IKBS users and suppliers, and the understanding of their needs and ambitions. Among possible demonstrator projects are:

4.4.14 Some less ambitious demonstrator projects should be chosen to be brought to fruition early in the timescale. However, even the simpler projects will be designed as complete developments subject to realistic constraints and systematic test use. The demonstrator projects provide a focus, a time scale and a motivation for the various research activities; assist in the transfer of know-how from researchers to producers of IKBS; and promote the education of IKBS users, especially by involving them in the design of systems for their purpose.

Relationship with other elements of the programme

4.4.15 The R&D activity of the IKBS programme will benefit from the concurrent activities in software engineering and on the man-machine interface. More specifically there will be individual demonstrator projects involving collaboration on MMI, e.g. in speech processing, and on VLSI, e.g. in chip design. Further, R&D on machine architectures will link IKBS activity with other areas, and indeed will benefit IT generally.

Milestones

4.4.16 The IKBS programme is intrinsically exploratory. Detailed milestones cannot be fixed in advance. However, the following targets should be aimed for.

Build Up of the R&D Community and Infrastructure
  1. People

    The community currently has about 40 experienced research workers and perhaps 150 people altogether, including younger workers and workers in border areas. The current annual output of masters students with relevant training is perhaps 20, and of doctoral students say 10.

    In 2-3 years, the IKBS community should increase by about 50 per cent, chiefly by attracting interested workers from bordering fields; enhanced training programmes should also contribute to the supply of younger workers. In 5 years the community should have at least doubled, with more experienced workers and an increased output of doctoral students.

  2. Support Facilities

    In 2-3 years'+ time an effective common base in suitable programming languages (e.g. LISP, PROLOG) and in useful programming support environments. The provision of adequate, appropriate computing machines and the communication net should be accelerating in 2-3 years software testing and production time, and the spread of know-how through the community.

Research Programmes

In 2-3 years' time increased understanding of:

  1. the basic capabilities required for IKBS in:
    • internal system capabilities in, for example:
      • the representation of knowledge by network and frame structures,
      • the use of knowledge in approximate inference and in the induction of new inference rules,
    • external system capabilities in, for example:
      • the recognition of continuous speech,
      • the interpretation of images,
      • the sensing of objects.
    • the ways in which types of IKBS tasks can be executed, for example:
      • database questioning recognising user motivation,
      • teaching basic skills by identifying pupils' mistaken rules.
    • the needs of particular application domains in, for example:
      • scientific diagnosis
      • circuit design
      • equipment maintenance.
By the end of year 5:
  1. Further understanding of capabilities, tasks and application domains.
  2. Substantial progress in understanding the nature of an IKBS as a whole, and of the means of integrating the modules of which it is made up.

    Thus we can expect a better understanding of IKBS capabilities for deduction within a changing body of knowledge and in the identification of topics in texts, in the conduct of IKBS tasks like program drafting, and in the needs of IKBS application domains like industrial process control. Also we can expect to be able to deploy knowledge and processes of different kinds and at different levels within a single system, for example, a robot with different types of input and output and a hierarchy of processors.

  3. Significant advances in the development of relevant technologies, both conventional and non-conventional, in particular in logic programming, and considerable experience in the use of parallel processors.
  4. A supply of useful packages such as expert system shells, natural language parsers and low-level vision processors. (This could be produced by years 2-3.)
Demonstrator Projects

In 2-3 year's time a range of modest expert systems along with some rather more ambitious IKBS could have been built, eg:

By the end of year 5 IKBS which are more ambitious and therefore more difficult to construct, eg:

Demonstrator projects exploit research results available at the time they begin, so demonstrators in the later part of the programme would consist of both:

  1. incremental versions of earlier systems, for example an adviser with voice input or robot with better vision, and
  2. systems which were radically better replacements of earlier ones made possible by research breakthroughs during the programme, for example a robot with full 3D as opposed to 2.5D perception.

By the end of the full 10 year programme even the most difficult demonstrator projects should have been implemented, if only in rather basic form. More importantly progress in the research programmes will have produced some IKBS demonstrators significantly more powerful than any available today, for example in robotics and in hardware and software design.

VLSI

4.5.1 Very large scale integration (VLSI) enables an order of magnitude increase in the number of interactions between circuits as compared to connecting LSI devices together on a printed circuit board. The performance and cost advantages this gives are vital to future IT products. The goal proposed for this part of the programme is to ensure that by the later 1980s the UK has secure access to internationally competitive VLSI. This will require the capability to specify, design, make and test silicon chips approximately one cm square containing approximately one million logic gates each capable of switching delays down to 1 nano second. That goal will not be reached in one step and en route products of lesser, but still competitive, performance will be generated. Other areas of work, such as speech and picture processing for man/machine interface and advanced processors for intelligent knowledge base systems, will require the development of VLSI demonstrator chips during this programme.

Computer Aided Design

4.5.2 Progress in VLSI depends as much on the management of the design complexity as on development of the silicon technology. Existing design tools are far from adequate to exploit the full potential of VLSI and a substantial programme for computer aided design is proposed. This is closely analogous to the software design process. Many aspects, particularly at the higher levels, will need development in the closest collaboration with the software programme. The objective is a set of tools which translate a system description into three outputs:

  1. A geometrical definition for reticle manufacture
  2. Test programmes
  3. Process definition.

The improved tools should cover:

For maximum use of these tools and the designs generated by them there is also a need for a database and file management operation.

Silicon technology

4.5.3 These CAD tools will be used by system designers to design advanced circuits. An indigenous capability to produce the silicon chips is also necessary and this must use world class silicon technology. Access to the capability is needed by all relevant firms although comparatively few firms will want to make the considerable investment in silicon process plant. Ideally firms building systems should have access to at least two sources of competitive but compatible silicon processing.

Range of processes

4.5.4 Several different processes will be needed

Together with work on interconnection and packaging.

For these processes a number of process techniques need to be progressed concurrently.

  1. Electron beam microfabrication
  2. X-ray lithography - whole slice or DSW
  3. Direct step-on-wafer (DSW) - projection aligners for high yield fine-line processing
  4. 'DRY' etching - using combination of plasma and ion bombardment to replace wet chemistry
  5. Ion implantation
  6. Epitaxial growth - whether on silicon, sapphire or other insulating substrates
  7. Layer processing - metals and dielectrics on the silicon surface, for contacts, interconnections, device delineation and passivation
  8. Annealing/dopant drive-in.

Process understanding

4.5.5 In support of the wide range of potential processes and process steps, it is important that there should be improved scientific understanding of the complex physics, chemistry and metallurgy involved in silicon technology. This calls for improved theoretical understanding, leading to better physical models for process and device simulation, and also for continuing improvement in the range of analytical tools available to study the behaviour of both functioning and non-functioning devices.

Using the results

4.5.6 This is a significant programme, the silicon technology aspects of which need to be backed by significant investment by industry in production plant. Systems designers will need access to the CAD. The present DOl CAD/CAM scheme is well structured to support the setting up of designer work stations interconnected as necessary by the proposed communications network.

4.5.7 The VLSI programme includes most of the elements of the MOD VHPIC programme although items very specific to MOD requirements such as technologies for radiation resistant components or demonstrators aimed at particular defence requirements are not included or costed. MOD have indicated that they would want to integrate the total programme, with the exception of demonstrators aimed at specific defence requirements, and have it managed as an entity.

4.5.8 VLSI programme targets

Year 1
  1. Establish limitations in currently available design tools and languages; define objectives for first generation of improvement and commence work.
  2. Order CAD work stations preferably funded through DOI CAD/CAM scheme and commence installation and interconnection through communication network.
  3. Investigate individual process steps for use in 1-micron bipolar and MOS processes, including improved emitter-base structure and pattern delineation at 1-micron feature size. In parallel, carry out basic device physics studies to understand behaviour of sub-micron devices and define the key characteristics of responsive processing.
  4. Commence component characterisation and data-base filling at 2-micron level to enable initial demonstrator projects to be designed. Identify target specifications for these chips.
Year 2
  1. Achieve and evaluate improved design aids for automatic layout of gate arrays; evaluate improved logic and circuits simulator. Complete design of improved two-dimensional device simulator incorporating second order physical effects.
  2. Complete CAD work station installation.
  3. Establish yield data on multi-layer interconnection technology at 2-micron feature size; complete development of tape automated bonding for chip assembly; continue device physics programme on sub-micron devices; evaluate test structures on 1-micron MOS and bipolar processes.
  4. Test first prototype of initial demonstrator chips. Commence design of further chips; discuss new specifications in support of MMI and IKBS programmes.
Year 3
  1. Commence development of improved behavioural specification and simulation language able to cope with complexities for the 1990s and adequate testability. Complete development of first generation of VLSI design tools.
  2. Evaluate first generation of 'responsive processing'; complete characterisation of 1-micron MOS and bipolar digital processes. Commence yield evaluation. Develop electron beam processing down to 0.5 micron for fast turn round process.
  3. Commence demonstrator chip design on 1-micron process. Continue creation of demonstrators at 2 microns.
Year 4
  1. Complete design of second generation CAD tools. Study new needs arising from new architectures, and benefits of IKBS to CAD.
  2. Continue device physics research, including exploration of new high density storage structures. Complete yield improvement programme on 1-micron processes and associated characterisation for design data base. Assess reliability and yield of 1-micron multi-layer metallisation process.
  3. Complete 2-micron demonstrator chips, concentrate on improved performance, higher complexity, 1-micron chips.
Year 5
  1. Develop improved process, device and circuit simulation tools to cope with the behaviour of sub-micron devices. Incorporate benefits of IKBS more fully into design aids.
  2. Evaluate novel process steps, such as photolytic impurity fixation. Complete development of responsive processing at 2-micron level and extendable down to 1 micron. Develop any special packaging techniques required by the demonstrator chips. Continue sub-micron research and comparison with other, non-silicon, technology.
  3. Maintain flow of 1-micron demonstrator chip designs, responding to the needs of the MMI and IKBS programmes.
Years 6-10

Take stock of the Total VLSI position - design aids, technology and products. Identify continuing needs for improved yields, performance, quick turn round for small numbers etc. Evaluate extrapolated silicon development against the then available alternatives (e.g. optical devices, III - V heterojunction bipolar IC's, Josephson technology etc).

COMMUNICATIONS

4.6.1 Good communications will be absolutely vital to the success of the AIT programme. The need for the programme in the first place derives in large measure from the fragmented state of the UK Information Technology research community, the components of which need to be brought out of isolation to work together effectively. Moreover, it is already clear that the success of the R&D activities will depend to some extent on breaking down barriers between hitherto separate areas of technology.

4.6.2 The objective is therefore to link participants together by means of a network, to create a new community, to promote interchange of news and information and to support collaborative working of scattered groups. Participants should be able to link up to this network, regardless of their local communications technology, from the start of the programme. As they will come to depend heavily on the network, it must be based on proven technology, rather than on experimental implementations.

4.6.3 It is therefore recommended that:

4.6.4 The need is foreseen for very high bandwidth communications between collaborating organisations to support a number of projects. The second generation IPSE is an example where graphics information will have to be transmitted between system components which may be sited remotely from one another. To anticipate such needs a programme of research and development should be initiated in the first year.

Table 5 Communications programme milestones
Year
1 Network set up as central communications vehicle for the AIT programme
2 substantial access via LANs
development of R&D communities, relying on network facilities
3
4 introduction of very high bandwidth facilities for (e.g.) 2nd Gen IPSE projects
5 wider use of very high bandwidth facilities
6-10 provision of very high bandwidth 2nd Gen network
⇑ Top of page
© Chilton Computing and UKRI Science and Technology Facilities Council webmaster@chilton-computing.org.uk
Our thanks to UKRI Science and Technology Facilities Council for hosting this site