Multi-agent Systems for Flexible Manufacturing Systems Management”. In From Theory to Prac

Multi-Agent System for Flexible ManufacturingSystems ManagementKrzysztof Cetnarowicz,Jarosław Ko´z lakInstitute of Computer ScienceAGH-University of Mining and MetallurgyAl.Mickiewicza30,30-059Krakow,Polande-mail:cetnar@.plAbstract.The paper focuses on the application of a multi-agent system for amanagement process.The presented system is working on the structure of graph,where nodes represent decision modules and edges–technological processes.Kinds of agents working in such environment are presented and an overview ofthe interaction protocols which can be used in such a system is given.1IntroductionThe development of production systems introduces new tasks,such as optimal produc-tion systems designing,efficient and safe production systems structure management or management of optimal functioning of a production process.On the other hand, proper management of a production system structure is very important for the enter-prise rentability and common problem.The contemporary production system becomes more and more complex and must beflexible for modifications.It makes a management process very difficult,inefficient and expensive.It seems that these tasks should be supported in their realization by computer sys-tems,or even totally managed by them.It may be realized when management systems are provided with models of a production process realized as decentralized systems. Attempts at realizing such systems improving the process of production systems func-tioning can be found in[7].An optimal movement of processed objects in the production system depends on a proper structure of the production system layout and real time routing,which can make use of the current production system structure in an optimal way.A proper choice of the production system layout structure has a basic influence on the effectiveness of activity. Consequently,it seems that multi-agent systems can be used as a model for a more complex computer management by means of effective object(parts)routing,but the evolution of the production system layout must be steered dynamically,thus adapting its structure to the current needs.2Principles of functioning of decentralized system managing computer networkWe may determine two goals of the computer management system for the production process[13],[1],[9],[5](fig.1):522nd International Workshop of Central and Eastern Europe on Multi-Agent Systems –management in real time to obtain the optimal functioning of the production pro-cess,–management of the production process structure(layout)to be adapted to the pro-duction realization.Fig.1.Schema of a computer system(with multiagent model)for production process manage-mentTo develop a multi-agent managing system for the production process we have to take a particular point of view to consider the production process as a graph environment for the agents population.In general,the production system may be represented as a network with boxes representing technological operations such as:milling machine, lathers,drills,grinder,etc.with a common operation:transport of partsfig.2.The parts enter the system,are treated and leave the system at the output.Fig.2.Schema of the production system layout consisting of milling machine,lathers,drills,heat treat,grinders,gear cutting and assembly as an output with circulating parts1,2,3,4.So the environment of that production system could be considered as a directed graph created in the following way(fig.3):–The production system is composed of a technological operation and edges repre-senting the operation,that follows.–The transport of parts between operations is considered as a technological opera-tion.–A node is placed between every technological operation.CEEMAS’01,26-29September2001,Krak´o w,Poland53–The whole production system layout may be mapped to a graph called a precedence diagram.–The edges represent technological operations,the nodes-decision nodes linking following operations and enables to take decision to switch the way of a part and select a proper path in the system.–The part is treated in the system following a properly chosen path in the graph.Fig.3.Schema of the production system layout-a,corresponding precedence diagram-b.The optimal treatment of a given part may be considered from two points of view:–point of view of the part:use the cheapest or the quickest way of operation execu-tion with a given quality standard,–point of view of the system:prepare the best adaptation of the system to perform demanded operations(be competitive)or use the cheapest way of the system mod-ification to adapt it to the treated parts needs(demand).The optimal management may be expressed in two ways:–Production management system functioning is based on the essential principle that defines conditions of a sending treated parts between the nodes following the opti-mal path(that is created in real time)of the part treatment.–The production system layout must be dynamically adapted to the treatment needs in an optimal way.These two points are dependent between them and must be considered as an in-tegrity during the management process.The main principle of the multiagent management system functioning is that the management process bases on the marked oriented decisions methods.To make a pos-sible application of the marked oriented management,it is necessary that every part pay for the technological operation executed on it.It means that a part pays for transfer542nd International Workshop of Central and Eastern Europe on Multi-Agent Systems through every edge between the nodes.Therefore a part is considered as a special agent (Part Agent-PA).Its task is to go by the graph from the source node to the destination node in given conditions and follow a path corresponding to the demanded technologi-cal operations[2].The conditions(the quickest,the cheapest treatment,etc.)are known to a given agent PA and are a basis for realizing a routing process of choosing dynamically the proper path.The payment may be done–directly:For the purpose,each node(exactly-output port on an edge)has an accountopened.Appropriate sums may be paid into the account by the treated part,Each processed part which is to be treated in the system has its account andtransfers an appropriate payment for sending to an account of a sending portfor the treatment by the corresponding edge.Agents authorized to use the founds accumulated on accounts may perform alloperations on the accounts with the intermediary of network operations.–indirectly:In a simplified case,a sending port obtains a generalized payment in a formof some number of(stipulated)points.The number of points collected makesevidence on activity of a given port,and what follows-a given operation(treat-ment)corresponding to the edge in a graph.The points collected may then be changed into cash and transferred from thefunds,(e.g.centrallyfixed),planned to be spent on production system func-tioning or development.The points collected may then be used as a basis to take decisions concerningthe optimal functioning and development of production system.The payments make a basis for market oriented decisions related to modifications (extension,decrease)of certain parts of a production system.Decisions are undertaken and then realized by the multi-agent system managing the production system.3Multi-agent system managing the production systemThe production system is dynamically mapped into the model(of the form of graph). The model is managed by the multiagent system.The multi-agent system consists of two main parts:a population of different kinds of agents and the environment in which the agents work.The environment is determined by the model of the production sys-tem,and may be considered as a particular graph(Fig.3).The model is created and dynamically updated to take into consideration the circulation of the treated parts in the system.There are the following types of agents in the system:–Input Part Agent(IPA)that provides the interface between the external reality and production system and creates the Part Agent.–Path Agent(PthA)that provides the proper path selection for the treated part.–Path Messenger Agent(PMA)that collects and transfers information around the graph about available operations(treatments)in the system.CEEMAS’01,26-29September2001,Krak´o w,Poland55–Port Agent(PoA)that makes possible for agents to leave a given node to reach the neighboring one by the selected edge,and to be treated by an operation represented by the edge.–Part Agent(PA)that represents a given treated part in the system,and provides a transfer of a given part from the beginning to the end following the proper path(i.e.sequence of operations)in the system.–Linker Agent(LA)that collects necessary information around the graph(or its part)to enable detection of a problem of traffic in the graph(such as overloaded operations,waiting lines for operations etc.).–Investor Agent(IA)that enables to undertake decisions with investment in the graph(extension or reduction of connections,new operations etc.)to modify the production system layout.–Investor Proxy Agent(IPxA)that participates in negotiations concerning invest-ments in the graph.Fig.4.Schema of the graph node where the agent undertake an action to establish new itinerary continuationThe real time optimal system management consists of a properly composed path for the treated part navigation in the production system ing its own required sequence of operations Part Agent goes across the graph from the beginning node to the ending node.At every node it enters the where cooperating with PthA,PMA,LA agents it obtains information about the possibilities of the path con-tinuation.Accordingly to its need(cheapest,quickest)it selects the best continuation of the itinerary and goes to the appropriate port where it pays(cooperation with PoA agent)for entering the following edge in the graph(which means that the corresponding operation will be executed on the part)The decisions concerning investments in the production process layout(which re-sults in the layout modifications)are settled as a result of negotiations among Investors and Investor Proxy agents(IA,IPxA).We can consider that the edge linking node with the node is overloaded(it means that the technological operations-drills be-tween and has insufficient efficiency rate).A scenario of the action undertaken by agents to establish a new connection(or to raise operation rate)may consists of the following steps(fig.5):–Port at the node linking it with the node is overloaded and the agent AoP corresponding to the port creates the LA agent that is sent to the node.562nd International Workshop of Central and Eastern Europe on Multi-Agent SystemsFig.5.Schema of the graph where the agent undertake action to establish new edges(operations, treatments)–At the node the AL agent verifies the traffic linking the node with the node.–If the traffic from the node to the node going via the node is too heavy, the agent Linker goes back to the node with a proposition to create a direct link between the nodes and,or looks for the next overloaded node(for example node).This proposition is paid to the Investor agent at the node.–Investor agent at the node builds afinancial project of the enterprise,and sends the Investor-Proxy agent to the node to negotiate its participation in the project.–If the negotiations are completed successfully the agent Investor-Proxy realizes the connection(direct or complex)placing order for the appropriate service with the appropriate company.4Interactions in the multi-agent system for the production system managementAn interaction protocol is a set of rules of interaction that describes what action each agent can take at each time.In this chapter typical interaction protocols used in the multi-agent systems are presented,then a presentation is given how these protocols can be applied to the interactions in the system for the production process management, which was presented in the previous chapter.CEEMAS’01,26-29September2001,Krak´o w,Poland57 4.1The interactions protocols used in the multi-agent systems.Contract net.Its idea is based on the mechanism of contracting,used in business[12]. There are two roles of the agents in the protocol:manager–agent,which orders the realization of the task.and contractors–the agents–potential executors of the task.Voting In some situations the agents can hold a vote on certain decisions.The result of the voting is obligatory for all agents participating in the voting.In[11]some protocols of voting are presented:plurality protocol,binary protocol and Bord’s protocol.Auctions The exchange of goods between the agents may be realized by auction.The agents participating in the auction can perform one of the followings roles:auctioneer or participant of the auction.There are many different protocols of auctions which are associated with particular strategies of the agents.On the basis of an overview from[10] we can enumerate:English(first-price open cry)auction,Vickrey auction(second-price sealed-bid),all-pay auction.Market interactions The interactions based on the market approach have several advan-tages,such as well analyzed(in economic science)theory and a low cost of communica-tion.There are two kinds of agents in the system:producers and consumers.Producers transform the goods using technologies they have and the consumers put demands for the goods needed and their decisions are undertaken on the basis of their utility function [15,4].The market approach was used for creation of systems strictly connected with the management of the graph or network:–the system for analysis of the distributed multi-commodityflow problem[14],–finding the optimal localization of the mirrors of services in the Internet[8],–system of routing of the packages in the Internet based on market mechanisms[6].4.2The types of the interactions in the multi-agent system for the managementof the product ion process.Introduction This sub-chapter will analyze what interactions protocols can be chosen for particular tasks of the production management system.We concentrate on the fol-lowing problems:–choosing the Port by PA agent,which enables the best quality of the connection (using the criterion based on the cost and/or the performance of the treatment),–gaining information about the state of the graph,–adding or removing the links to/from the graph by Investor and Investor Proxy.Choosing the Port by Part Agent The process of choosing the Port by the Part Agent may be realized using the following interactions protocols:–Contract Net–PA agent informs all Ports on the Node that it wants to get a defined treatment(which means that it is looking for a particular node in the graph).The Ports inform what price it has to pay for the available treatment,and what are other properties of the treatment(time elapsed,quality etc.)The PA agent chooses the best from its point of view of treatment–it means the corresponding Port.582nd International Workshop of Central and Eastern Europe on Multi-Agent Systems –voting–The PA agents vote which treatment is the best and the corresponding edge gets a privilege to be selected for such treatment.–auction–If the production system is heavily loaded(overloaded),the PA agents can take part in the auctions of treatment offered by the Ports(i.e.edges).The opposite situation is also possible–the agent PA demands to buy the treatment offered by one of the Ports at a given node.There are more than one available ports offering similar kinds of treatment and the Ports participate in the auction offering lower and lower prices and better treatment.The PA agent chooses the lowest price, the best treatment.This algorithm may be used if the production system is currently not loaded.–market–The complex negotiation of the cost of the treatment between PA agents considered as consumers and Ports playing a role of producers can be implemented using a market model.Collecting the information about the current state of the network.Linker Agent can gain the information about the state of the network using the following interaction pro-tocols:–Contract Net–LA announces that it is ready to buy information about the state of the network.The group of the nodes(Node Managers)can send their proposals concerning the quality and the costs of the possessed information.–Market–LA and Port Agents participate in the exchange of the information about the structure of the network.The prices of the information are set by the market methods.The modification of the structure of the production process structure(graph)-adding or removing the operations(edges)The modification of the structure of the network by adding or removing the edges representing technological operations can be performed using the following interaction protocols:–Auction–Agents Investors take part in the auction of adding a new edge to the graph or selling the one by the Investor who actually owns it.It is also possible to increase or reduce the efficiency,productivity or quality of a given operation(edge).–Market–The Investors can buy or sell technological operations(edges).5Simulation of the multi-agent system for the production system layout(graph management)The multi-agent system proposed for the production system layout management has been modelled and verified by simulation(fig.6).It has a limited function–provides only the migration of the Part Agents using the Contract Net Protocol.Other tasks,such as collecting information about the structure of the network and changing the structure of the network(adding/removing nodes)are not realized.The following interaction among the agents has been realized:CEEMAS’01,26-29September2001,Krak´o w,Poland59Fig.6.Schema of the examined production system layout-graph corresponding to the system:a -at the beginning of simulation,b-at the end of simulation after the improvement of the system layout(extension of operations-corresponding edges)–Part Agent when arriving at a given node tries to reach the destination node by the edge with demanded operation at the lowest costs.–At the node a number of options of transfer continuation at different prices are proposed to the Part Agent.–Part Agent accepts the solution,taking into consideration the price of the following operation versus its own resources and the quality of the treatment versus the time it can waste for all the operations from the beginning to the end of the part treatment.–Part Agent pays the Port Agent for the treatment.With the resources saved the Port Agents(and Node Manager)undertake modification of the production system layout(graph).The simulation of the system was performed on the computer network composed of 6nodes with edges representing technological operations.Technology needs and pricing policy favour the following edges of the graph: (B,F),(B,G),(C,G),(D,G),(E,F),(E,G),(F,G).The initial state of the network with the same productivity rate of all edges(technological operations)is shown infig.6a. After a number of operations execution and transfers of parts in the system the nodes using collected funds modify the parameters of the network.The result of the system investment is presented infig.6b.After the improvement(changes of the efficiency of operations corresponding to the edges)realized by the system the overloaded operations (edges)were reinforced(thick lines in thefig.6b).6ConclusionThe proposed system of a automatic management of production system enables both optimal real-time management of the production system functioning,and optimal,self-acting adaptation of the network structure to users’needs,also in real-time[3].Both aspects of optimal management,considered together may bring very interesting results. The application of the above mentioned system will lower costs of production system exploitation.In the future it will enable the realization of optimal management of pro-duction systems layout for large,extensive production systems.The proposed system602nd International Workshop of Central and Eastern Europe on Multi-Agent Systems enablesflexible production systems management(of graph structure)with the use of economic criteria of optimization.The system takes advantage of possibilities given by the decentralized,multi-agent system,that enables making local decisions,taking into account local,and if necessary and possible-a global point of view.The presented system may be easily extended to uniform,optimal management of network,including not only hardware management but also other resources,-work software.References1.K.Cetnarowicz,E.Nawarecki,and M.Zabinska.M-agent architecture and its applicationto the agent oriented technology.In Proc.of the DAIMAS’97.International workshop:Dis-tributed Artificial Intelligence and Multi-Agent Systems.St.Petersburg,Russia,1997.2.K.Cetnarowicz,M.˙Zabi´n ska,and E.Cetnarowicz.A decentralized multi-agent systemfor computer network reconfiguration.In J.Stefan,editor,Proc.of the21st International Workshop:Advances Simulation of Systems-ASSIS’99.MARQ,Ostrava,Czech Republic, 1999.3.Krzysztof Cetnarowicz and Jaroslaw Kozlak.Multi-agent system for decentralized com-puter network management.In Proceedings of the Second Conference on Management and Control of Production and Logistics(MCPL’2000),Grenoble,July2000.4.John Q.Cheng and Michael P.Wellman.The WALRAS algorithm:A convergent distributedimplementation of general equilibrium putational Economics.,1997.5.N.Gaither.Production and Operations Management.The Dryden Press,New York,1994.6.M.A.Gibney and N.R.Jennings.Dynamic resource allocation by market based routingin telecommunications networks.In Garijo F.J.Albayrac S.,editor,Intelligent Agents for Telecommunications Applications(IATA’98).Springer Verlag,1998.7.G.Dobrowolski K.Cetnarowicz,E.Nawarecki.Multi-agent mrp class system for productionand disassembling.In Management and Control of Production and Logistics.IFIP,IF AC, IEEE Conference,Grenoble,France,2000.ENSIEG,LAG Grenoble,France2000.8.Tracy Mullen and Michael P.Wellman.A simple computational market for network infor-mation services.In ICMAS.1995.9.J.F.Proud.Master Scheduling.John Wiley&Sons Inc.,New York,1994.10.Tuomas W.Sandholm.Negotiation among self-interested computationally limited agents.PhD thesis of University of Massachusetts Amherst,1996.11.Tuomas W.Sandholm.Distributed rational decision making.In Gerhard Weiss,editor,Multi-agent Systems.A Modern Approach to Distributed Artificial Intelligenc,volume II,chapter7, pages201–258.The MIT Press,1999.12.Reid G.Smith.The contract net protocol:High-level communication and control in a dis-tributed problem solver.IEEE Transactions on Computers.,C-29(12):1104–1113,1980. 13.W.J.Stevenson.Production/Operations Management.Richard D.Irwin Inc.,HomewoodIL60430,1990.14.Michael P.Wellman.A market-oriented programming environment and its application todistributed multicommodityflow problem.Journal of Artificial Intelligence Research,pages 1–23,1993.15.Michael P.Wellman.Market-oriented programming:some early lessons.In S.Clearwa-ter,editor,Market-Based Control:A Paradigm for Distributed Resource Allocation.World Scientific,1996.。