Optimization and multi-agent control in manufacturing processes.

Hrubina, K. ; Sebej, P. ; Hrehova, S. 等

Abstract The paper deals with the possibilities of a multi-agent control application as well as with design and function of a decision making system based on the utilization of mathematical models simulations provided by PC. The paper also deals with the properties of a multi-agent system and its application to the processes control in a complex system.

Key words: processes optimization, decision making process, multi-agent system, multi-agent system application

1. INTRODUCTION

Nowadays, in new economic environment, enterprises and private firms have to solve the problem of productivity of labour and its rentability. Sustainable technological development makes special demands on enterprises and companies.

In addition, the situation of companies is hindered by quick changes on markets as well as ever-growing customers requirements towards the quality of selling products. That is why the competing ability and economical development of a company or enterprise are becoming dependent on an enterprise information and control system.

It is necessary to realize that information and control system is a set of workers, procedures and sources which collects, transforms and distributes information or facilities providing the organization control in order to achieve the set aims. It is obvious that such a system is created and developed with the help of computer-aided facilities and methods of information technology. That is why we emphasize the principal interconnection between information and information flows and the processes providing the manufacturing control and the development of an enterprise or company in accordance with the actual strategic goals.

In principal, we deal with the optimization trend from the point of view of time and economy. From the above it follows that the actual task is the solution of effective manufacturing planning problems and effective time table which will provide optimal utilization of manufacturing capacities and will considerably increase the productivity of labour. Prompt and flexible satisfying of customers orders connected with the manufacturing are the other indicators of a modern enterprise.

In order to solve such complicated optimizing tasks in practice, the suboptimal solution that enable to reduce manufacturing costs, increase of manufacturing capacity as well as its flexibility will often suffice. The task defined in this way is the task determined for the multi-agent system of decision and scheduling of operations.[Maoik, 2003; Doran, 1992; Hrubina et al., 2005].

The aim of the paper is to present the conception of the integrated system of enterprise management and characteristics of its individual levels as well as to present the decision-making system. The authors also deal with the possibility of application of the multi-agent principle to the selected subsystems control within the enterprise.

2. ON DECISION--MAKING SYSTEM DESIGN

First of all we are going to define the terms "decision" and "strategy". Under the term "decision" we are going to understand the determination of values of input parameters in the given stage of a controlled process. Under the term "strategy of decision" we are going to understand the sequence of the step-by-step decisions. Strategy which satisfies the conditions of the preset defined criterion of optimization will be defined as the optimal strategy.

Based on the block diagram "Hierarchy of the Control Functions with Vertical Decompositions,[Jadlovska, 2004], after the operational unit "Adaptation", the operational unit "Knowledge of Processes" will be placed.

It stands to reason that such a system has at its disposal inputs, outputs and at the same time it is affected by external noise. In such a case, under the term "process" (regarding the first paragraph of the paper) we understand the whole net of the manufacturing enterprise which includes manufacturing, distribution and communication with a customer.

With such division of the decision-making process we can effectively observe the state of each participating unit as well as to predict its behaviour, to treat the assumed and presumable noises and to enforce the flexible maintenance of a separate nodes of manufacturing enterprise net.

In the conception of the integrated system control, the term "functional levels (or "layers") of control" comes to the fore.

In general, we deal with the following functional levels (layers) of control:

* direct control level,

* optimization level,

* adaptation level,

* the level of system knowledge

* organization level.

The fourth level of the system knowledge integrates the wide basis of theoretical methods and empirical procedures obtained from the process behavior under investigation as well as experience of the whole control system and manufacturing technologies, Fig.1.

The expected advantages of the above mentioned hierarchical approach are as follows:

a) Improved problem analysis

b) Methods and algorithms adaptation

c) Software conception

d) Adaptability

e) Control levels

Automated system of production control can be described and investigated as any complex system from different points of view. This system includes manufacturing and technological process of a controlled system. It can be completely described only by a set of elements of mutually coherent structures. They are the following structures:

* functional, algorithmic, organizational, information and technical.

3. ON MULTI-AGENT SYSTEM

3.1 General Description and Properties of a Multi-agent System

The definition of a "agent" has not been stabilized yet. Definitions presented in literature can be divided into two categories: general definitions and special definitions for particular environment, e.g. computer environment, mechanical agents, analogue agents, etc. In this paper under the term "agent" we are going to understand an independent unit able to receive inputs and using its own or accepted strategic procedure (within the definition of a "strategy") to affect environment and itself. Production equipment of technical, biological or information character are usually considered to be the agents. The present-day understanding of the concept "agent" in its narrower sense as program systems is not telling. According to their strategy procedure, the agents have the attributes of autonomy, communicativeness, co-operation, ability to negotiate, etc.

[FIGURE 1 OMITTED]

If the agents are grouped to the sets regardless of the strategy of their (multi-agent systems) creation as well as regardless of the heterogeneity of the individual agents, their mutual or isolated activity in the common environment, cooperation or antagonism, then we are going to define these sets as the multi-agent system, [Marik et al., 2003].

In general, the multi-agent system is an inhomogeneous set of agents with the particular goals and operative mutual relations as well as relations towards the environment. Each agent is provided with the tools in order to achieve the particular goal in the given environment. To achieve the particular goal, each agent creates conditions for the best and quickest obtainment of the solution using the forms of cooperation and influence of other agents activity. [Dobrowolski,1997].

Each agent with respect to itself is autonomous and its actions and future conditions are dependent on its preceding conditions as well as environment condition, i.e. other cooperating agents are also considered as a part of environment.

Individual activity of agents is synchronized only in accordance with the stimuli that arise in the environment as well as the conditions which will occur. If general synchronization exists, it is considered to be the imposed phenomenon. Asynchronous activity is the typical feature of the multi-agent systems. Based on natural mechanisms of interactions, activities show the synchronizing effect.

The multi-agent system is considered to be the open system to other agents which can finish their activity if they influence the process of solution toward the optimum.

Communication among agents is characterized by controlled approach towards actual and historical information enabling the agents to create suppositions for the selection of more effective strategies as well as other mechanisms, e.g. the mechanism of learning.

3.2 Application of the Multi-agent Principle to the Selected Processes Control within the Enterprise.

Nowadays, the notion agent is used in the decision making area, planning, and production management.

In general, the term "agent" also covers the independent program system (within the computer environment) which is able to make decisions autonomously and based on these decisions to perform activities (commands) the realization of which leads to the goal achievement while the whole process is carried out with the optimal tools utilization. To activate the agents in the subsystems of the presented complicated system we are going to present some algorithms created by the authors of the following works: [Hrubina--Jadlovska, 2002; Hrubina et al. 2002; 2005]. Their program realization was performed on the basis of the mathematical models used in the area of projects management, production planning, thermal processes control as well as within the tasks of the systems with distributed parameters optimal control using programming systems MS-Excel, QSB+ and MATLAB

1. Algorithm of the CPM (Critical Path Method) method and the PERT (Program Evaluation and Review Technique) for projects management and its program realization based on MS-Excel and the QSB is presented in the work: [Hrubina et al., 2005].

2. The Simplex algorithm and its application to the tasks of production planning and management is presented in the work: [Hrubina -Jadlovska-Hrehova. 2002].

3. The algorithm for solving the task of optimal control of heating the material in the furnace where the thermal process is expressed by mathematical model with the defined boundary and initial conditions and loss minimization expressed by the integral criterion, is presented in the work: [Sebej--Hrubina--Ragan, 2005].

4. CONCLUSION

The paper deals with the problems of processes and complex systems control. The achieved results are also presented in the process of investigation. The paper also presents the complex systems hierarchic control with the description of the function levels of control as well as the decision-making system design. From the theoretical and practical viewpoints, the suppositions for the possibility of multi-agent system control application were created.

5. REFERENCES

Dobrowolski, G. (1997). Network Operating Agents as a Mean for Decentralised Decision Support System, In: Conference on management and control of production and logistics MCPL'97, Cmpinas--SP--Brazil, pp. II/498-503

Doran, J. (1992). Distributed Ai and its Applications. Heidelberg, Springer Verlag, 617 p.

Hrubina, K.--Jadlovska, A. (2002). Optimal Control and Approximation of Variational Inequalities Cybernetics. The International Journal of Systems and Cybernetics. MCB University Press of England. Vol. 31, No 9/10, pp. 1401-1408, ISSN 0368-492X

Hrubina, K.--Jadlovska, A .--Hrehova, S. (2002). Methods and Computer Aided Solutions of the Operation Analysis Problems. Informatech, Ltd. Kosice, 325 p., ISBN 80-88941-19-9.

Jadlovska, A, et al. (2004). Optimal Control of Complex Processes. In :Annals of DAAAM for 2004. Vienna. DAAAM International, p.177-178. ISBN 3-901509-42-9

Hrubina, K. et al (2005). Optimal Control of Processes Based on the Use of Informatics Methods. Kosice, Informatech Ltd., 287 p., ISBN 80-88941-30-X

Marik, V.--Stepankova, O.--Lazansky, J et al (2003). Artificial Inteligence, (1), (2), (3), (4) Praha, Academia.

Sebej, P.--Hrubina, K.--Ragan, E.,(2005). Optimalization of Thermal Processes Utilizing Mathematical Model on PC. (In Slovak), In. Manufacturing Engineering, TU v Kosiciach FVT so sidlom v Presove, ISSN 1335-7972