Figure 1 - Intelligent agents work
(Animation consists of 6 frames, 21Kb delayed by 0.2 - 0.3 seconds between frames, the delay before repeat is 1.25s; reproduced cyclically)
Faculty: Computer Science and Technology
Speciality: System Programming
Simulation is currently the most affordable, convenient, fast, safe way to verify a new product from design to start the process of exploitation.
Network dynamic objects belong to the family of objects that are difficult to comprehend fully without tools (such as modeling). Their condition is constantly changing linearly with time, the physical dimensions are too great, any erroneous operation can lead to significant costs, accidents and even loss of life, so the simulation of network dynamic objects (SDO) is a very important problem, whose solution requires the analysis, study, research, implementation approaches to the creation of reliable, fast and user-friendly models.
Agent-oriented programming approach can parallelize the process of modeling, with the speed of calculation. On Agent Based approach, the computation goes to a higher level, because agent - a standalone program that interact with other agents to obtain a single solution to the problem. The task can specify the user and the agent himself, which is able to plan their further actions on the basis of belief bases, strategies, calculations, analysis of information obtained - ie a process becomes more autonomous and the Intelligent software.
Network dynamic objects can no longer be considered objectively, the environment of sovemennogo rights. The actual trend of the modern world is informatsionngo accelerated integration and merger of large dynamical systems. Hence, the design, calculation and simulation of reliable, fast LMS systems must take first place in the activities of people.
ARS is relevant and promising trend in modern programming as well as in turn, was the PLO in the days of structured programming. Agent-oriented approach allows us to move to a higher level programming simulators network objects with distributed parameters, to introduce elements of intelligence in the AMS.
Aim is to develop a parallel simulation modeling of automated systems management of network dynamic objects using agent-oriented approach, as well as the development of the system automation.
Main objectives of research and development are:
Scientific innovation is a new way of automation systems, network dynamic objects with the help of agent technology, and this approach can be used at different levels and components of automated systems of dynamic objects.
Parallelization and distribution of algorithms of information systems by traditional methods has many serious weaknesses and limitations:
- need to overcome the boundaries of operating systems, due to the fact that in distributed systems use different operating systems, protocols and Interfaces;
- diversity of object models is manifested in the fact that classes and objects, constructed in various instrumental settings, have some differences;
- the complexity of managing distributed customers, especially in variable network configuration;
- methodological limitations associated with using different models and methods for constructing the components of distributed systems.
Distributed object architectures (CORBA, Java RMI, DCOM, WEB-services) seek to overcome these limitations, but leave unresolved the following problems:
- the need to recompile the program code when changes are made in the objects and interfaces;
- can not dynamically adapt the behavior of software objects, depending on the state and behavior of the environment;
- inability to work explicitly with the models of knowledge, the value of which is constantly growing;
- the accumulation of huge amounts of information which can not be semantically processed and presented in a form suitable for the perception and processing.
To solve such problems require new approaches for systems development. In this connection there is the paradigm of agent-oriented systems using intelligent agents as a high-level abstraction for formalizing and structuring the domain and as a powerful tool for designing and implementing complex information systems.
Intelligent Agents - a new class of software and hardware and software entities that act on behalf of the user to locate and process information, conduct negotiations in electronic commerce systems and services that automate routine operations and support to tackle difficult tasks, cooperate with other software agents in case of complex problems, thus removing the redundant information to the human strain.
Basic properties of agents are considered [1]:
- autonomy - the ability to perform activities independently;
- homogeneity / heterogeneity - the ability to combine homogeneous or heterogeneous functions;
- trainability and intelligence - machine learning, behavior correction to improve its own performance;
- active behavior - constant exchange of information within the agent and with the environment;
- communication - data exchange with other agents;
- perception of the environment - the presence of special means of perception of the environment functioning agent;
- mobility - the movement of the agent inside other software and physical environments.
Agent-Oriented Approach (AOP) has some fundamental differences from the already traditional object-oriented programming (OOP) [1]. The agent is a more complex, active and autonomous unit (Fig. 1).
Figure 1 - Intelligent agents work
(Animation consists of 6 frames, 21Kb delayed by 0.2 - 0.3 seconds between frames, the delay before repeat is 1.25s; reproduced cyclically)
The PLO computational process is understood as a system, built from modules that interact with each other and have their own ways of handling incoming messages. A turn, AOP refines this framework by fixing the activity of the modules, agents and changes in their states through the analysis of beliefs, commitments, intentions, etc.
The presence of the agent mechanism of goals provides a new level of autonomy [10]. The agent does not necessarily available to any other agent or user, but just depends on environmental conditions, including the goals and intentions of other agents. In contrast to the object, the agent can take on certain obligations or, conversely, to refuse to perform certain work, claiming lack of competence, employment by another task, etc. At the same time, the agent can perform actions such as generation, removal, change of other agents, the intensification of their functions, scenarios, activities, remembering the current state of other agents.
All this clearly indicates that the agent, being an "active object", which forms its own behavior, is at a higher level of complexity in relation to the traditional objects in the PLO.
Modeling complex dynamic systems in most cases be an obligatory component of agents management and control of networked dynamic objects. Power modeling allows the agent to have a mental picture of its structure the external world - an object that is being examined, controlled. Thus an agent or combination of agents can expect and anticipate the consequences of their actions with the object, correcting errors and their intentions [3]. Block modeling for distributed network systems may also comprise agents with information about your area of the object. Agents coming together to jointly address the problem, the calculation of flow, the air pressure in various parts of the ventilation network.
To develop agent-oriented systems there are special libraries and components are for the most common programming languages to help in the creation and establishment of a developed system of agents.
Most famous projects: AgentBuilder, Bee-gent, Cable, Decaf, FIPA-OS, Grasshopper, Gypsy, JADE, JASON, JAFMAS, MAML, ProcessLink, Swarm, Zeus.
Software requirements can be divided into two classes: system and instrumental technology. The first class should include: full support for all phases of the life cycle of agent-based system to have a single integrated development environment for all system components, support for different categories of users, provision of visual design, the automatic generation of executable program code of intelligent agents; support for collective work on the draft system and automated documentation of all phases of the development process. The second class includes components, such as ensuring the development of distributed knowledge bases and inference mechanisms, support the formation of logical models of distributed systems (forming skeleton of the areas and distribution of knowledge of intellectual components of agent systems), provision of methods for constructing models of agent behavior, implementation of mechanisms of parallel operation, Communication and coordination of agents, support extensibility models of agents in real operation, etc.
Rather wide application received multiagent programming platform JADE (Java Agent Development Framework), is fully implemented in Java. JADE simplifies the development of multi-agent systems through the use of tools, support phase configuration and deployment of the system. Agent Based Platform can be installed on computers with different operating systems, and configured via a remote GUI-interface even during the execution of agents.
For JADE platform developed Wednesday JADEX, which is an extension of multi-agent platform. This environment involves a hybrid reactive-deliberative architecture in which the agent is seen as "black box" which receives and sends messages. Based on results of message processing, internal and external events, deliberative mechanism to decide on the transition to a new plan of action or continue the old. The current plan can send messages to other agents, change the basis of opinion, create new goals and cause internal events. The system uses a library of plans, which are treated as Java-classes.
IDK supports two types of visual description: AUML language and methodology INGENIAS. AUML is an extension of language UML, specialized for the description of agent-based systems. INGENIAS methodology based on defining a set of metamodels that describe the behavior of each agent, the interaction between agents, the organization of multi-agent systems, environment, goals and objectives defined for each agent.
Methodology consists of the construction of these metamodels [9]:
- organizational metamodel, which describes the architecture of multi-agent system and its functions. Options are determined at the time of the job the organization's objectives and procedures of their implementation;
- metamodel environment, describing the nature, affecting the perception of agents;
- metamodel aims and objectives, describing the agent's state changes in time depending on specific tasks, achieving goals and actions to be performed if the goal can be achieved,
- metamodel agent, describing the behavior of each agent;
- metamodel interaction, describing the behavior of two or more interacting agents. Depending on the language describing it can be represented as a UML-collaboration diagrams or AUML-chart.
Obtained in the form of metamodels description of multi-agent system can handle the relevant modules. Behavior of agents is determined by their goals and objectives, as well as the interaction between them [3]. Goals, in turn, can be broken down into simpler goals that can be achieved certain sequences of tasks. The set of tasks and interactions determines the global behavior of the system. On the basis of the description of the system is generating code. The resulting description can be used to create the code of multi-agent system for a concrete platform.
Thus, all the above software products can help a developer of network systems modeling of dynamic objects with distributed parameters at the highest level of programming.
The work will be developed by the automation system by using DLS agent-based approach. As well as the related system modeling of the system. Results can be applied in the coal industry.
This work belongs to cooperate with the University of Stuttgart, and will help further the maintenance and development of scientific cooperation.
Theme simulation of dynamic objects has been well studied. In DonNTU written set of master works, theses, in one way or another review, the development of those simulations.
However, it must keep pace with the times and move to address challenges set by the world society, namely the introduction of agent technologies in scientific research, and start the main goal - to study the problems of implementation of the parallel simulator environment and this issue will help understand agent technologies
Search Results Full theme final work in the world have shown that development using agent technologies are extremely rare within the CIS and Ukraine (for example, Russia is developing for the analysis of economic and market the use of AOP). In the United States and European countries give the matter much attention. But all the same theme of master rabotyfakticheski is unique and pretend to novelty in the scientific world. Result links on specific aspects of topics see links.
The work will be developed by the automation system by using DLS agent-based approach. As well as the related system modeling of the system. The results can be applied in the coal industry
This work belongs to cooperate with the University of Stuttgart and will contribute to further plan the maintenance and development of scientific cooperation.
The paper reviewed the basic principles of agent technologies, the need to use this approach to automated system management of network dynamic object
In Ukraine there are many articles on the study of the use of agent technology. In this work it was proved that new technology is very promising because it allows to move to a new level of engineering and programming, to increase reliability, speed and interactivity tools for modeling complex dynamic systems.
At the time of writing the abstract Master's thesis was on the stage of development. This summary may differ from the final version of the abstract in the master's work.