Introdution

In the last decades several techniques and systems have been developed for dealing with the increasing problem's complexity, in terms of its formulation and resolution. Such systems, however, present several difficulties in which computational ones are primus inter pares. In this context, Distributed Computation can be used as a technique and an archetype for the integration of Artificial Intelligence techniques and Distributed Computation, giving birth to a new discipline, Distributed Artificial Intelligence (DAI). DAI was initially defined as a kind of problem solving where the data (knowledge) and computation (reasoning) are logically and/or physically (geographically) separated [Nilsson-1981] [Davis-1980].

It's only natural that this new discipline integrates with Multi-Agent Systems. Added value must be considered, such as co-operation, co-ordination, rationality, intelligence. Whenever alternatives exist for the execution of a task, and this task is to be carried out by rational entities (i.e., agents), there's a need for grouping. I.e., agents must have social interactions [Neves e Machado-1997]), arguing abilities and negotiation techniques (e.g. the Contract Net Protocol [Davis e Smith-1983]).

Multi-agent Systems allow for strategies of non-hierarchic distribution of tasks. The solution is to create a community of agents co-operating with each other to the resolution of the problem, which assumes some form of social organisation [Analide e Neves-1997].

In the Manufacturing arena these systems gave origin to Intelligent Manufacturing Systems (IMS) and Holonic Manufacturing Systems (HMS). IMS give us a vision of future [Van Brussel-1995]: "The Intelligent Manufacturing System takes intellectual activities in manufacturing and uses them to better harmonise human beings and intelligent machines integrating the entire corporation from marketing through design, production and distribution, in a flexible manner which improves productivity". IMS will possess the innate ability to respond, promptly and correctly, to changes in requirements. They differ from conventional manufacturing systems - even advanced ones - in their inherent capability to adapt to changes without external intervention.

In what respects to HMS, they can be seen as dynamical and decentralised systems, which naturally integrates Human beings. HMS are based on the notion of Holon [Koestler-1967]. The Holon is a part and at the same time the whole, i.e., an Holon can be made of other Holons, and at the same time is part of other Holons.

Figure 1 - Holonic Architecture

The concept of Holarchy appear (alternatively to hierarchy), as a society of holons which co-operate to fulfil a certain goal. The holarchy defines the basic rules of co-operation, restricting their autonomy. A brief introduction to Holonics and a glossary of terms can be found at hms.ifw.uni-hannover.de/public/Concepts/fr_conce.htm.

In Manufacturing Systems Resources (machinery) can be considered as holons, which group themselves into cells and/or Production Lines (which constitutes another holon). There can also exist logically (i.e., without physical existence) defined Holons such as the Process Planning Holon, or the Production Planning Holon. In a scheduling system there can be holons representing Resources and Tasks [Sousa e Ramos-1998].

Objectives

This work will develop communities of holons which help solving problems in industrial environments, using specialisation and decentralisation of tasks [Neves e Machado-1997].

The system will have the ability to learn how to react in extreme situations or exceptions (e.g., machine breakdown, rush order, etc.) according to past performance (e.g. case based learning or hypothesis generation).

References

(c) 1999, Paulo Sousa
comments
Created: October 8th, 1999
Last updated: November 13th 1999