Multiagent Systems - Research Proposal Example

MULTI AGENT SYSTEM TABLE OF CONTENTS 3 Introduction 4 An overview of Multi-Agent system 4 Multi-Agent systems modeling in Eco Tiles factory 6 Different agents in Eco Tile factory and its purposes 8 Advantages of Multi Agent system in Eco Tiles 10 Non-multi agent approaches 11 Conclusions 12 References 13 Abstract The aim of this work is to apply multi agent system technology to model a real production programming problem in Eco Tiles, a fictitious ceramic tile factory. The production programming is regarded to be the main problem in Eco Tiles because obtaining a reactive production planning and scheduling system is inevitable in manufacturing. This work presents a multi agent system modeling for production planning and scheduling problems in Eco Tiles where constituent agents cooperate to find a possible schedule considering factory layout, capacity, time constraints and on-line orders. The multi agent system looks at improving the production performance, increasing the schedule reliability and efficiency and keeping updated schedules. Introduction Software agent technology, which started in the early 1980s as a sub-field of artificial intelligence and has been a very dynamic field of research for about more than two decades, focuses its research mainly on aspects of multi-agent planning and distributed problem solving. Even though the agent technology research field is still heterogeneous it has set up itself as a major stream of computer science and artificial intelligence in particular. Multi-agent system is quite imperative for a ceramic tile company like Eco Tile factory, because it needs dynamic production processes to offer online programming to the customers and it would be able to manage real time response about their services and delivery times of required products through the process where design and other activities are done ahead of the production process. Like most other production units, production programming was one of the main problems of this company. The multi-agent system in Eco Tile turns to be a better structure for dealing with design and development of an application which should be flexible and adaptable to the environment. Multi agent technique seems to be appropriate for solving complicated problems that require intelligence. Here, Multi agent system provides a natural way to solve problems that are inherently distributed within the factory. An overview of Multi-agent system Recently, multi agent system shows its trend towards the overt design and organization use where heterogeneous agents work together within distinct roles to attain individual and general goals. When emphasizing on team goals, organizations allow agents to work together with the help of individual agents to function properly and to perform the jobs that they are best appropriated for. When focusing on individual agent’s goal, organizations allow agents to perform collaborative tasks with other agents for the benefits of each agent. “A multi-agent system can be considered as a loosely coupled network of problem solver entities that work together to find answers to problems that are beyond the individual capabilities or knowledge of each entity” (Durfee, Lesser, & Corkill, 1989). “Multi agent system is a collection of agents” (Wooldridge 2004). An agent is a software process that comprises of autonomy, social ability, pro-activeness and reactivity. Autonomy agents can operate without direct or indirect involvements of humans or others. Social ability does interact with humans and other agents through some sorts of communication language. Even though Pro-activeness does not normally interact with their environment, it’s capable of taking initiative and exhibiting goal directed behavior. Reactivity agents recognize its environment and respond in a timely manner. Basically, Multi agent system has two differences; firstly, the objects are passive and hence it reacts to external stimulus and it does not show any goal directed behavior. Secondly, agents generally use a common communication language between agents but object messages are class dependent. How to model and implement individual and social behavior in distributed systems is the main sector where multi agents system investigates. It looks at ideas and notions like autonomy, reactivity, perception, and goal directed reasoning so as to model and apply individual behavior. Same time, it explores aspects of co-ordination, co-operation, synchronization, negotiation, role assignment and self organizations so that it can create social behavior. Multi agent system technology provides a better and convenient platform for modeling workflow. The environment, where each agent is represented as an accurate information unit of the enterprise which is modeled, is quite suitable model for maximization and visualization of material, information and workflows. Apart from workflow management, the multi agent technology is best suited for manufacturing processes simulation, production control, planning and scheduling. Here, the enterprise management necessarily does not need to replace the information system which is currently running by entirely a new technology. It makes use of technology that is there already. Existing software system manages and maintains organizations processes like resource planning, material management, human resource management etc. Multi Agent system modeling in Eco Tile factory Eco Tile factory in is a domain for applying multi-agent technology because the increasing development level achieved by this multi agent technology allows undertaking with complicated problem fields and meanwhile this system needs software application. The production programming problem is regarded to be one of the main critical issues in Eco Tile factory as in any other ceramic tile factory. Multi agent system has been modeled attempting to simplify to maximum environment conditions. The concerned environment is very dynamic and it reflects the dynamic conditions and constant changes in Eco Tile factory, such as new customer requirements, and availability of machineries etc. In Eco Tile factory, the master structure can be represented as major input data to turn out the production scheduling. This generation process may be described by weak automatiization in which a number of schedules are static and each one is a simple manual conversion. The process is also not reactive, where scheduling systems do not face the set of different things occurred in a period of break down or supplier fault. The schedules are based on global models that are not considering both peculiarity of each level of production process and differentiated stage of each production. The production planning department of Eco Tile factory sends a planning lot to the multi agent system so that it can schedule and model it. The multi agent system queries the warehouse system so as to get raw materials availability. The multi agent system automatically checks out the availability of raw materials in the warehouse. The multi agent system queries the warehouse system so as to get raw materials availability. The multi agent system automatically checks out the availability of raw materials in the warehouse. From the Eco Tile factory plant system it gets updated data of the plant status and machine condition of its functionality. And finally, the schedule which was defined or modified is sent to the production manager in order to carry it out in the factory plant. See the figure below. Dynamic Production Planning in Eco Tiles From the Eco Tile factory plant system it gets updated data of the plant status and machine condition of its functionality. And finally, the schedule which was defined or modified is sent to the production manager in order to carry it out in the factory plant. From this application it is very obvious that the multi agent system controls four situations. a. The schedule creation situation which is done and activated by production planning. b. The schedule execution monitoring so as to identify any possible factory failures c. The schedule modification in order to reschedule machine tasks and to fulfill production goals d. The management of the organization. Different agents in Eco Tile factory and its purposes 1. Managing agent. It is the regulatory unit which controls and regulates the cooperation among the different agents of the system. It also finds out any failure in the creation or modification of the schedule that needs to be communicated to the production planning department. 2. Scheduling agents group. Scheduling group is set up by four agents; one coordinator agent and three scheduling level agents. These stage agents are characterized in creating partial schedules for each production stages of glazing, kiln and classification. Each stage agent performs different suitable algorithms under heterogeneous circumstances. While glazing agent focuses on the problem of high sequences set up costs in the production lines, the Kiln agent tries to minimize changes of its temperature curve. The coordinator agent looks at obtaining practicable production schedules for the system as a whole while taking in to account the tasks proposed by lot planner agent. 3. Schedule controller. This agent’s task is to manage the information about a new schedule order getting from production planning department. It looks at resource assignment for a specific planning lot. It not only creates a new schedule from planning lot but also cooperates with the scheduling group for allocating jobs to resources. It is possible that there can be more schedule control agents running at a time but each of them has the task of creating specific schedule order without lapsing time. 4. Schedule modification agent. This agent manages information regarding changes required for adjusting the schedule due to any possible failures or mistakes in the manufacturing processes. These data and information are recognized from plant status data. 5. Lot planning agent. This agent’s role is to maintain the job sequences and concerned information to manufacture a lot effectively. In short, the Multi agent system provides a suitable framework for dealing with production programming in Eco Tile factory and it in fact reflects the dynamic conditions within the factory. Advantages of Multi Agent system in Eco Tiles Like any other project oriented and manufacturing enterprise Eco Tile factory also faces the problems of how to establish the optimal production plan much effectively and how to attain the highest turn over rate by utilizing company’s resources maximum. To avoid these crisis of Eco Tile factory and to keep business running more effectively by receiving maximum number of production orders, an extra attention needs to be paid to the production planning of Eco Tiles. So for Eco Tiles, when all other systems fail, multi agent system could be a choice as a standard solution. More over, agent technology allows planning algorithms to be distributed throughout the system, hence the algorithms planning keep on working on real time and hence the overall planning is reprocessed immediately when new job comes in. Production control itself is attractive to Eco Tile factory as the production, which is the important industrial activity within Eco Tiles, is inherently distributed and dynamic system exhibiting many changes during the operation. Agent technology is capable of meeting the forthcoming challenges in production control. The Eco Tiles production turns to face a fundamental change from seller’s market to the customer’s market. Customers demand constant product innovation, low cost customization and better services. Customers always choose the product which meets their requirements best. With the help of multi agent system, Eco Tile factory would be able to increase product variety, satisfy customers’ demand quickly, and to reduce investment costs. Non-multi agent approaches There are objections to multi agent systems as they are having similarity with other fields. For example, Distributed or concurrent system where synchronization and coordination are not structured at design time because they are in distributed system. In multi agent system of Eco Tiles factory, the agents are self interested while components in a concurrent system have the common goal of maximizing the overall efficiency of the system. Due to these reasons, multi agent system forms a better framework to Eco Tile factory than that of distributed system. Another non-multi agent approach is Artificial intelligence which mainly focuses on the study of intelligence components like learning, understanding, planning etc, meanwhile the goal of research about agents is the integration of all these elements. Suppose, if artificial intelligence is implemented in Eco Tile factory, the social ability in systems would be ignored but it is imperative in an intelligent behavior. Another non-multi agent field is Economics or Game theory where the concepts are descriptive and it does not indicate any thing about how to compute a solution. Game theory is built on the basis of rationality ideas. In economics or game theory system the agents are rational while programming this rationality is the main problem in multi agent system Conclusion The paper presents system architecture to manage the production programming problem in Eco Tile factory. The architecture and methodological development is based on the multi agent system paradigm as it provides a natural and clear way to solve problems in manufacturing industries. The main goal consists of reaching an order management system in Eco Tile factory which would be able to satisfy client’s requirements. It is necessary to integrate different distributed production steps in flexible, versatile and natural way in order to satisfy the goals. The paper provides an overall structure of multi agent system with brief descriptions of different agents and its functions. It aims to obtain a system that improves the current production programming processes in Eco Tile factory, so that it can make the production more reactive, dynamic and automatic and can enhance its production system and customer services. References ALLAHVERDI A., GUPTA J. & ALDOWAISAN F. (1999) A review of scheduling research involving setup considerations, Omega BEATRIZ LOPEZ (2005), Artificial Intelligence Research and Development, IOS Press DURFEE E.H, LESSER V.R & CORKILL DD (1989), Trends in cooperative distributed problem solving. IEEE Transactions on Knowledge and Data Engineering GERHARD WEISS (1999), A Modern Approach to Distributed Artificial Intelligence, MIT Press MICHAEL J WOOLDRODGE, STEFAN BUSSMAN and NICK JENNINGS (2004), Multi agent Systems for Manufacturing Control: A Design Methodology, Springer. MICHAL PECHOUCEK, ALES RIHA, JIRI VOKRINEKI, ExPlanTech: Applying Multi-agent Systems in Production Planning, Gerstner Laboratory, Department of Cybernetics, Czech Technical University MULTI AGENT SYSTEM, Retrieved 18/9/08 from http://www.multiagent.com/ SCOTT A DELOACH, Engineering Organization- Based Multi agent system, Multi agent and Cooperative Robotics Laboratory, Department of Computing and Information Sciences, Kansas State University Read More