"Education is not a preparation for life, it is life" (J. Devey)

Modern organizations, processes and computer systems which components are people and/or difficult-to-formalize objects are complex and distributed. Traditional formal ways to describe these automation objects do not provide adequacy of created models.

Multi-agent systems theory is a new information technology paradigm that combines scientific and technical achievements of artificial intelligence methods, modern local and global computer networks, distributed databases and distributed computations.

Multi-agent systems theory creates new quality models of difficult-to-formalize objects in a form of intelligent organizations presented by autonomous artificial agents. Both multi-agent methodology estimation and created models quality evaluation are important.

The purpose of work is multi-agent simulation of industrial mixtures production process as the example of complex distributed system.

Following tasks should be solved:

1. Review of multi-agent methodologies and development tools.
2. Syntax and semantics description of agent models.
3. Multi-agent modelling of industrial mixtures production process.
4. The program agents architecture selection.
5. Multi-agent system development tool selection.
6. Multi-agent system of industrial mixtures production process program implementation.
7. Multi-agent modelling results analysis.

Consist in:

• the development of new agent-oriented programs representation way;
• the modelling quality improvement due to the intellectualization of artificial agents behaviour and interaction;
• the creation of flexible system for the analysis, control and re-engineering of industrial mixtures production process.

The conventional artificial agent definition has not been established yet.

According to FIPA specification (Foundation for Intelligent Physical Agents is the international organization responsible for standardization process of multi-agent systems and technologies), an artificial agent is the entity placed in some environment, able to observe it, to receive different data, to interpret them and to execute commands effecting the environment. The artificial agent can contain program and hardware components [5].

So the artificial agent is understood as a high-level abstraction for formalizing and structuring difficult subject domain concepts. It is presented as a hardware-software feature able to operate autonomously to reach own goals [3].

The intelligent agent is understood as active, autonomous, sociable, reasoned entity operating in difficult, dynamic and more often virtual environments [3].

There is a set of intelligent agent’s properties – autonomy, social behaviour, reactivity, pro-activity, basic knowledge, beliefs, desires, goals, intentions and commitments [1].

The multi-agent system can be defined as a collection of interconnected program and/or hardware agents, capable to co-operate with each other and with the environment. Agents possess certain mental abilities and can operate individually or together. Tasks in multi-agent system are distributed between the agents, which co-operate for the general problem solving.

There are three basic classes of multi-agent models and architectures types:

• deliberative architectures and models – the architectures based on knowledge processing methods;
• reactive architectures and models – the architectures based on behavioral models if a form of "stimulus-reaction";
• hybrid architectures and models.

Methodologies of agent-oriented analysis are applied at the analysis and design stages of multi-agent system development. There are four main agent-oriented analysis methodologies classes:

• based on the object-oriented methods and technologies with appropriate add-ins (AUML);
• using traditional methods of knowledge engineering (MAS-CommonKADS);
• based on the organizational-oriented representations (Gaia [3, 8]);
• combining methods of three previous classes.

Development tools are used at the implementation and testing stages of multi-agent system development. They use their own models to detail multi-agent systems. There are two main classes: frameworks (JADE [7], Agent Development Kit [2]) and development environments (Zeus [6], Agent Builder).

The object of multi-agent simulation is a real technological process of industrial mixtures production [3].

Objects (devices) and subjects (staff) are distributed territorially, possess difficult behaviour.

The multi-agent model should provide necessary adequacy to real technological production process.

Participants of production process (devices and staff) are modelled by autonomous program agents of the appropriate architecture. Each agent possess powers and behaviour of the participant it represents.

By co-operating with other agents and evaluating a current state of production process, each agent independently makes a decision about its further actions.

Successful solution of the industrial mixtures production process multi-agent simulation allows:

• to model the production process under normal working conditions;
• to evaluate and reorganize production process when failures occur;
• to evaluate the efficiency of production structure.

Representation of industrial mixtures production process participants in the form of program agents allows to create flexible system model and to easily re-engineer the production structure and estimate its viability.

At the beginning the agent-oriented analysis of a subject domain was made. As a result of such analysis abstraction and conceptual models of multi-agent system were created. These models were used then for further specification at the development tool level. Organizational-oriented methodology Gaia and Zeus development environment were selected.

The main concept of Gaia is role. Gaia methodology creates a set of models: roles, interactions, agent, services and acquaintance models (fig. 1).

During the agent-oriented analysis several roles have been defined [3].

Zeus development tool operates such base concepts as agent, fact, purpose and task. It creates following models: ontology, program agents, tasks, coordination and organization models (fig. 1).

Figure 1 – Gaia methodology models and Zeus development tool models interrelation

The method to transform Gaia methodology models into Zeus development tool concepts was proposed, the results are represented in the structural and algebraic forms [4].

Interaction between two agents of multi-agent system was examined [4]. Here what is done:

• role schemata were created (as a part of Gaia roles model);
• protocols of those agents interaction were described (as a part of Gaia interactions model);
• part of Gaia agent model was created;
• part of Gaia services model was described;
• part of Gaia acquaintance model was created;
• developed Gaia models were transformed into Zeus toolkit concepts, ontology and agents descriptions were received;
• program code in Java language was generated in Zeus toolkit.

• further possibilities research of multi-agent simulation of assigned task;
• further possibilities examination of chosen multi-agent system development tool;
• program prototype development of industrial mixtures production process multi-agent system;
• modelling results analysis: efficiency evaluation of multi-agent methodology application for solving the assigned task, quality estimation of the created models.

The master’s work has not completed yet. Completion date is December 2010. Full text can be received from the author or his scientific advisor after this date.

1. Городецкий В.И., Грушинский М.С., Хабалов А.В. Многоагентные системы (обзор) [Электронный ресурс] / В.И. Городецкий, М.С. Грушинский, А.В. Хабалов. – Режим доступа: http://www.raai.org/library/ainews/1998/2/GGKHMAS.ZIP
2. Зудикова Ю.В., Федяев О.И. Разработка программных агентов в инструментальной среде Agent Development Kit / Ю.В. Зудикова, О.И. Федяев // Комп’ютерний моніторинг та інформаційні технології – 2009 / Матеріали V науково-технічної конференції студентів, аспірантів та молодих науковців. – Донецьк, ДонНТУ. – 2009. – с. 272-274.
3. Зудикова Ю.В., Федяев О.И. Разработка многоагентной модели процесса производства промышленных смесей / Ю.В. Зудикова, О.И. Федяев // Інформатика та комп’ютерні технології / Матеріали V міжнародної науково-технічної конференції студентів, аспірантів та молодих науковців – 24-26 листопада 2009р. – Донецьк, ДонНТУ. – 2009. – с. 261-264.
4. Зудикова Ю.В., Федяев О.И. Трансформация моделей методологии Gaia в концепты инструментальной среды Zeus при многоагентном моделировании процесса производства промышленных смесей / Ю.В. Зудикова, О.И. Федяев // Інформаційні управляючі системи та комп’ютерний моніторинг (ІУС та КМ-2010) / Материіали I всеукраїнської науково-технічної конференції студентів, аспірантів та молодих вчених – 19-21 травня 2010р., Донецьк, ДонНТУ. – 2010. – с. 196-200.
5. Швецов А.Н. Агентно-ориентированные системы: от формальных моделей к промышленным приложениям [Электронный ресурс] / А.Н. Швецов. – Режим доступа: http://www.ict.edu.ru/lib/index.php?id_res=5656
6. Hyacinth S. Nwana, Divine T. Ndumu, Lyndon C. Lee. ZEUS: An Advanced Tool-Kit for Engineering Distributed Multi-Agent Systems [Электронный ресурс] / Hyacinth S. Nwana, Divine T. Ndumu, Lyndon C. Lee. – Режим доступа: http://www.agent.ai/doc/upload/200302/nwan98.pdf
7. Giovanni Caire (перевод с английского Зайцев И.М.). Руководство для начала работы с JADE [Электронный ресурс] / Giovanni Caire (перевод с английского Зайцев И.М.). – Режим доступа: http://www.masters.donntu.ru/2009/fvti/zaytsev/library/book1/
8. M.Wooldridge, N.Jennings, D.Kinny. The Gaia methodology for agent-oriented analysis and design [Электронный ресурс] / M.Wooldridge, N.Jennings, D.Kinny. – Режим доступа: http://www.csc.liv.ac.uk/~mjw/pubs/jaamas2000b.pdf