Agents and Agent Based Design Approaches To Engineering Design and Manufacturing

1
Agents and Agent Based Design Approaches To
Engineering Design and Manufacturing

• A. Wallace C. Boldyreff
• RISE
• Department of Computer Science
• University of Durham

2
Format of Presentation

• Briefly overview CARD project.
• Define agents, define agent characteristics and
  how agents have been used.
• Outline agent frameworks.
• Discuss agent based design.
• Discuss the benefits and limitations of using
  agents.

3
CARD Project

• Is an EPSRC/SEBPC funded project being carried
  out with British Steel.
• Aimed at supporting British Steel in their move
  to centralise a previously independent design
  function consisting of distributed designers,
  legacy design knowledge and design records.
• Integrate these with new knowledge sources using
  agent technology into a centralised function
  which allows steel production to be optimised
  across mills.

4
(No Transcript)
5
(No Transcript)
6
Diagram Showing Rolls For Steel Production
7
Goal of Roll Design

• The ultimate goal of roll design is to acquire
  the desired geometry and tolerance of a steel
  product with the least cost whilst preventing
  defects in the workpiece.

8
Definitions

• An agent is a computer system situated in some
  environment, that is capable of autonomous action
  in this environment to meet its design objective.
  (Jennings and Wooldridge 1998).
• An agent is software that assists people and acts
  on their behalf. It is delegated to perform some
  tasks and given constraints under which it can
  operate. (P.C. Janca and D.Gilbert. 1998).

9
Definitions Continued

• An agent is a component of software and/or
  hardware which is capable of acting autonomously
  to accomplish tasks on behalf of its user. (Nwana
  1996).
• However, Nwana suggests, it is an umbrella term,
  meta-term or class, which covers a range of other
  more specific agent types, which can be listed
  and defined respectively

10
(No Transcript)
11
Agent Classification Characteristics

• There are several dimensions along which we can
  classify software agents and agent-based design
  systems. The following few slides provide an
  overview of the current state of the art in agent
  development and how we might classify agents.
• The categories are not mutually exclusive and
  different types of agents have different
  combinations of these characteristics.

12
Classification of Agents
13
Application of Agents
14
System Classification
15
Framework of Agents
16
Environment of Agents
17
What Do Agents Offer?

• Ability to solve problems that have, so far, been
  beyond the scope of automation, either because
  no existing technology is suitable or an existing
  solution is too expensive.
• Have the ability to solve existing problems more
  efficiently than current solutions.

18
Current limitations of Design Support Systems

• Difficult to integrate the number of isolated
  design systems together for a large design task,
  despite each package being well developed for its
  task.
• Resolution of conflicts.
• Integration of systems and agents across
  different hardware and operating systems.
• Providing adequate search methods to support
  decision making in large decision making spaces.

19
Limitations, continued

• Integration and co-ordination of symbolic
  reasoning, neural networks, numeric computations
  and graphics processing.
• Integration of intelligent design agents, design
  software and commercial AI tools and methods.
• Using existing network structures and techniques
  to support development of distributed knowledge
  based architecture, knowledge communication
  protocols and parallel computing algorithms.

20
Agent Based Design

• As opposed to designing Agent-based systems,
  this is defined as the process of designers
  developing designs using an agent-based system to
  support their activities

21
Agent Architecture in CARD

• Requirement is for a number of design agents to
  work together towards the same design objective.
  This requires the co-operation of several
  intelligent design agents which could represent
  the currently isolated knowledge sources and
  support systems.

22
Proposed CARD Agent Architecture

• An integrated distributed intelligent design
  system, such as the multi-agent system being
  developed by the CARD project, can solve these
  issues and problems. Such a system can offer

23
Proposed CARD Architecture, continued

• An integrated design platform
• A knowledge based communication protocol
• A supervising agent - Meta-Agent System
• A distributed computational architecture
• An intelligent user interface

24
Integrated Architecture Levels
25
Meta-system Integration
26
Meta-System Architecture
27
Benefits of Agent Based Design

• Agent approach can support
• Data control, expertise and resources which are
  inherently distributed.
• Where a system is regarded as a society of
  autonomous co-operating components.
• Where a system contains legacy components which
  must be made to interact with other, often new,
  components.
• Agent based approach provides a metaphor and
  framework for reasoning about design systems.

28
Limitations of Agent Based Design

• Agent-based solutions have been inappropriately
  applied
• Lack of overall system controllers
• Lack of global perspectives
• Requires designers to shift from application to a
  user centered paradigm approach
• Increased complexity of management in the
  net-centric nature of these systems
• Complexity and lack of coherence typical of
  engineering design activities requires a flexible
  agent architecture to provide support

29
Conclusions

• Need to address complex issues of application
  specificity of various agents and their
  interaction requirements
• Need to give agent-based systems a learning and
  adaptability component
• Formulation of principles for agent-based design
  required

• Final thought - agent technology looks
  increasingly promising as a basis of a new
  approach to engineering design focussed on an
  agent-based model.