Multi-agent Systems in Manufacturing System - Coursework Example

Multi-Agent Systems in Manufacturing Course Tutor Date Table of Contents Table of Contents 2 1.0 Introduction 3 2.0 What is an agent? 4 3.0 What's multi-agent system? 5 4.0 Characteristics of Multi-agent systems 7 4.1 Cognitive Characteristic 8 4.2 Interaction Capabilities 9 4.3 Multi-agent Organisation 11 4.4 Work Standardisation 13 5.0 Coordination and Cooperation 14 6.0 Coordination planning 15 7.0 Negotiation 17 8.0 Agent Architecture 18 8.1 Centralised Multi-Agent Architecture 19 8.2 Decentralised Multi-agent architecture 22 9.0 Multi-agent System Communication 24 10.0 conclusion 26 11.0 List of References 27 1.0 Introduction Multi-agent systems are tested and proven to create efficiency when it comes to dynamic scheduling in flexible manufacturing systems. Agents that make up a multi-agent system are characterised by autonomous behaviours coupled by organisation and a stringent pattern of communication. A multi-agent system is described by this article as an artificial system that is made up of diverse autonomous agents which are well collaborated and coordinated in order to achieve a major organisational goal. The behavioural characteristics of multi-agent systems are also analysed with intent of establishing the connection with real life scenarios. It is also evident that inasmuch as this system calls for operative autonomy, there is need for coordination and cooperation in order to achieve greatly. The types of multi-agent systems mainly covered by this paper are; centralized and decentralized. It is established that the major difference between these two systems is the level of autonomy and coordination. The centralized systems of multi-agent architecture purely involve a system where a single agent manages other autonomous agents towards the achievement of a single objective. Contrary to this, decentralized multi-agent architecture is characterised by completely autonomous agents which carry out their activities to achieve sole goals. These agents however interact with others within the system since they are bestowed with knowledge to plan on their own with regard to a common goal to be achieved by the system. This paper is motivated by a theoretical research on the working principles of a multi-agent system and its characteristic for purposes of deployment within flexible manufacturing systems to reduce energy consumption. 2.0 What is an agent? According to Ferber (1993) in Kouiss et al (1997), an agent is a virtual or entity which stands alone within a given technical surrounding. This may also be populated by other minor agents that seek to contribute towards the final action of the main agent. In order for an agent to perform its mandated actions, it is required to communicate with other agents of the same or greater capacities within the environment of coexistence. The resulting behaviour in agents is obtained from observations, knowledge or the interactions that guide the successful performance. Some of the identifiable features of an agent include the capability to represent the environment partially or by perception, ability to communicate with other agents, ability to produce son agents and the autonomy behaviour in which the objectives are preserved. An agent is split into three layers namely the static knowledge layer, expertise layer and the communication layer. A static knowledge layer as applied in the structuring of agents is said to contain knowledge within itself as a host to the specific memory that is meant to remember every single step that concerns its environment of existence. The expertise layer usually contains the treatment representation for a given knowledge action that is supposed to perform the intended action within a given time frame through logical expressions or algorithms guided by a certain set of irrefutable production rules. The communication layer identities all the communication tools that build up a successful communication protocol between one agent and the other. Depending on the form of communication that is set in place, this layer takes into account of all messages that are sent and received (Kouiss, Pierreval, & Mebarki, 1997). Figure 1: Structure of an agent (Kouiss, Pierreval, & Mebarki, 1997). 3.0 What's multi-agent system? Based on the understanding of an agent, a multi-agent system is an artificial system which is made up of diverse autonomous agents which are well collaborated in order to achieve a major organisational goal. These agents usually work simultaneously while pursuing individual goals that contribute to the achievement of a single final objective (Kouiss et al., 1997). Each independent agent involved in a multisystem is perceived in terms of its usefulness as a problem solver within a given environment due to the level of artificial intelligence that is bestowed upon it. The feasibility of an agent as an optimal problem solver lies entirely on the heterogeneity, hierarchy and structure as shown in the figure below (Lee & Kim, 2008). Figure 2: Multi-agent system layout (Nie, Bai, Wang, Liu, & Cai, 2012). Agents possess problem solving capability and are therefore meant to interact with each other via a given communication protocol in order to cooperate. Multi-agent systems interactions are further categorised into cooperative and self-interested interactions which depict the optimal course of action that a given solution takes. The synchronous and asynchronous communication protocols that are involved are meant to boost the accuracy of information relayed from one level to the other to allow for global planning. Comparing these complex interactions to real world situations, it shall be noted that the overriding resolution setups work in the same way of deployment (Andreadis, Klazoglou, Niotaki, & Bouzakis, 2014). The adaptability concept of multi-agent systems refers to the changes that have been monitored over time through trial and error in order to come up with a workable system. The proactive nature of taking initiatives by these systems exhibit a goal oriented comportment. Multi-agent systems thus act out of reasoning to adapt with their interaction patterns and hence end up being characterised by such notions as beliefs, goals and plans that develop over time. The optimal courses that are acted upon by agents making up a system are based in the preferences or goals that are set for the entire system. This means that the autonomy of a given agent may be affected by the rational choices that may be undertaken by another agent that runs concurrently in order to deliver a given goal. 4.0 Characteristics of Multi-agent systems Multi-agent characteristics are dependent on the basic properties attached to agents. Therefore, in order to create an understanding on characteristics of multi-agents it is mandatory to understand the basic properties of agents. To begin with, characteristics of agents are observed with regard to the environment, belief and knowledge, environmental state and actions related to environmental change. The environment within which agents act is observable and predictable although it is kept autonomous as stated in the definition section. The autonomous state comes with the ability to decide based on the available opportunities. The rationality of agents is said to be bound by computational complexity that is associated with the optimization objective and subsequent limitations (Monostori, Vancza, & Kumara, 2006). 4.1 Cognitive Characteristic Multi-agents are characterised by problem solving capability through inbuilt sensory mechanisms and decision on course of action to be undertaken on careful analysis. This information is then communicated from one agent to another for execution purposes. The problem solving faculty lies squarely in the searching, reasoning, planning and learning in that order. There is however a notion that power synthesis applies both symbolic and sub-symbolic methodology to carry out classical and quantitative decisions that are well reasoned and advanced in terms of applied models. The classical decision theory applies widely when it comes to making choices for the entire system although the alternative decisions may be expected from utility agents too. Data is always available in a multi-agent system to facilitate for quantitative and qualitative analysis prior to decision making (Monostori, Vancza, & Kumara, 2006). Knowledge is shared explicitly within a multi-agent system for goal and action determination. Goals are described as a set of objectives to be achieved thus they are the key component for a rational system behaviour. A multi-agent system is characterised by artificial intelligence which is the main drive of the planning methods which have been apparently approved to work in real-life situations. The BDI model of expression deployed by these systems is motivated by environmental knowledge. Based on this model B represents beliefs that are associated to the operational environment. This is usually expressed in simpler terms search as long term statuses and goals set to be achieved. D represents desires which are to be main motivations of the already set intentions (I) as a show of commitments that were previously made by the agents. Intentions have been found to stabilize multi-agent systems inasmuch as the environment of existence is thought to be dynamic. These may also be transformed into plans that are laid down to be achieved under well-coordinated circumstances. The cognitive state is therefore described by the widely publicised BDI model of problem approach which entirely depends on beliefs, desires and intentions. The cognitive nature is further exhibited by the ability of agents to continue updating themselves based on perceptions that emanate from environmental beliefs and desires (Shen, Distributed Manufacturing Scheduling Using Intelligent Agents, 2002). 4.2 Interaction Capabilities A multi-agent system is defined as a combination of autonomous agents that are meant to achieve a common goal (Baffo, Confessore, & Stecca, 2013). In order to achieve these goals, agents are designed in a way that they interact with each other directly or indirectly via the operational environment. Interaction mechanisms include coordination media and protocols designed for distributed planning. Coordination protocols are usually designed to offer control over agents’ interaction for a common decision to be reached. Contract net protocol (CNP) is widely used by monitor agent for the purpose of dictating what is to be done by subject agents and when. The contract net protocol is modelled based on the mechanism contracted by the organisation in exchanging services and goods. The main agent usually receives bids by subject agents to carry out a task once it is announced for execution. The evaluation is carried out based on task distribution and other real life factors such as the prices quantities, dates etc. Apparently, the coordination media is based on shared memory that is meant to relay communicated data in a seamless manner. Examples of this kind of media include pheromones and blackboards based on stimergy coordination mechanisms (Ferber, 1993). Due to the operational nature of agents, conflicts do arise from time to time. This calls for negotiation mechanisms to be enacted for arguments and negotiations to take place. Coordination ensures that tasks are not delayed by conflicts thus enact collaborated and planned actions which seek to achieve a common goal. Agent communication languages (ACL) rely on artificial intelligence of communication models which are developed to ensure that all communications are fulfilled. ACL is also described as performative due to knowledge in query manipulation ability that is assertive and manipulative. The consensual nature and interoperability of the agent community is an ontology whose specification is explicitly conceptual in terms of structure. The ontology concept is based on a logical language that distinguishes the desired qualities of communication. The consensual nature of communication enforces understanding among agents due to accumulated knowledge. These ontologies are also facilitative by nature in that they allow for automated machine processing through interoperability among agents (Lim & Zhang, A multi-agent based manufacturing control strategy for responsive manufacturing, 2003). 4.3 Multi-agent Organisation Multi-agent systems are characterised by organisation just as in real-life situations. There exist organisational patterns that are based on responsibilities such as collection and combination of results. The basic human patterns have been adapted for the purpose of organisation and subsequent streamlining of system operations. For organisation purposes the tasks to be conducted by the agents are subdivided into roles and relations. Roles are identified in accordance to agent rights and obligations so as to ensure smooth running. The task manager is designed so as to control task assignment, monitoring of execution and results combination. This entirely depends on the chosen system architecture for performance of various basic roles which add up in order to make up a major achievement. The admissibility of agent interactions depends on the organisational roles and protocol definition. The rules governing these roles are well coordinated through a well-designed mechanism that is adapted from existing organisational forms (Braga, Rossetti, Reis, & Oliveira, 2009). Figure 3: Direct supervision of agents in a multi-agent system (Monostori, Vancza, & Kumara, 2006). According to Monostori (2006) the task manager conducts a direct supervision over the operator agents that form up a system. The responsibilities taken by the task manager is well coordinated through a repeatable pattern with a hierarchical setup. This system works just as a real life setup that is controlled by a single manager. All information is channelled through the system task manager which then redistributed this information in accordance to capability and level of interaction. This form of interaction also reduces the occurrence of conflicts which would rather derail the whole process. For illustration purposes, system requests relayed by agent (R) and processed by agent (O) as the operator is handled by a managerial agent (M) and depicted in terms of (A) as shown in the flow diagram below. 4.4 Work Standardisation Multi-agent systems are also characterised by work standardisation as part of a delegate system where operators are instructed by task managers in accordance to task specifications. The execution part is however left to the agents to carry out at their own discretion. Object flow is controlled by operators who work in a decentralized and a rather randomized manner. A Coordination is further covered in section 7.0 in order to portray the arrangement of agents when it comes to expertise (Wang & Lin, 2009). Mutual adjustment is part of the coordination characteristic of a multi-agent system in which the managerial and operation levels are not distinguished due to inability of the agents to communicate. In such a scenario, the coordination method is simplified to cater for the appropriate commands within a set time frame. The cooperation levels in such a system utilising mutual adjustment, altruism is highly rated (Cetnarowicz & Kozlak, 2001). This is directly opposite of the characteristics of the negotiation methods that are described in section 5.2. Figure 4: Mutual agreement (Monostori, Vancza, & Kumara, 2006). Mutual adjustment technology of coordinating multi-agent systems thus deploys the convenient team formation and dissolution. According to Monostori et al. (2006) the human organisational models can be surpassed with the application of the multi-agent systems. The decentralization of problem solving functions are also used in computational models that are aimed at coming up with workable simulations for self organisation. In self organisation, the systems are adapted to a changing structure that can be attributed to requirements and environmental change. This does not however prevent the agents from coordinating for the good of the system. 5.0 Coordination and Cooperation The fact that a multi-agent system is formed of autonomous agents, there is a requirement to instil life into it through interaction. Due to the commonness of goals that are being pursued by the system, agents are required to behave in a rational manner that is likely to boost the progress. It is evident that a system that is distorted by conflicts cannot work effectively to achieve the main interests. It is therefore important for the entire system to be linked in a way that cooperation is achieved without straining (Guo & Zhang, 2010). Autonomous agents have however been found to be egocentric in human terms although they are characterised by further rational behaviour that aids in achieving the final goal as set by the system. High-level performance is however attributed to coordinating agents which are also entrusted for decision making. Agents are always aware of the consequences of egocentric behaviour thus are motivated by cooperation as an intelligent means of creating accessibility to important information which would rather benefit a single agent. The barriers that the multi-agent system incurs can be transformed into the strength that shall steer the system to yet another level of success altogether. In order to cooperate effectively, the agents are not only required to be proactive but also reactive. This shall ensure responsiveness is achieved in a timely manner through commitment and collaboration. Mutual commitment is deemed as a necessary in achieving joint goals while simultaneously achieving solely motivated ones (Shen, Distributed Manufacturing Scheduling Using Intelligent Agents, 2002). Cooperation, collaboration, and Coordination contribute majorly towards manufacturing processes. This is especially seen when it comes to the heterogeneous system of product manufacturing. For example the highest level of business functions is based on plans execution throughout the organisation. In order to achieve this goal, the 3C have to be enacted effectively within the organisation as to become the fabric that holds it. The value chain is also dependent on the 3C in that it affects the final product quality with the life cycle axis in mind (Sudo & Matsuda, 2013). 6.0 Coordination planning Coordination planning in assembly lines is important within the agent architecture study since they are based in collaborative systems of management. Process planning is carried out in a special mediation agent that subsequently assembles all the functions to be executed. The coordination function of this system is associated with database management as a way of transferring information from one centre to the other as the negotiator sets in to aid in the deployment of functions generated to various agents that are mandated for execution. In case where the process planner is involved, the coordinator is entrusted to transferring information to agents and of execution fails, the negotiator sets in to offer support by negotiating functions between the two agents as a means of goal execution for task delivery (Cetnarowicz & Kozlak, 2001). Figure 5: Planning in a centralised multi-agent system (Andreadis, Klazoglou, Niotaki, & Bouzakis, 2014). Alliance agents are involved in control of agents that are associated with the execution states for purposes of load balancing. This comes up with a registry in which the information coordination and execution by agents is monitored and controlled for effective achievement of set goals. The alliance agents engage in supporting the distributed design for analyses of factors associated with the assembly and product cost. The evaluation agent gets information through queries directed to the alliance agents which offer a great reasoning mechanism for joining and quitting of the multi-agent system. Evaluation knowledge is expressed to the system using an object oriented approach that provides the production-rule as the main technique of reasoning (Erol, Sahin, Baykasoglu, & Kaplanoglu, 2012). Coordination planning in decentralised multi-agent system involves a network of personal computers and servers which are dedicated to coordination. Generally the user interface that is provided for the purpose of coordination is the personal computer web browser. Agents in a defined local area network are maintained within their area of operation through the implementation of a firewall technology. The internet provides a leeway for communication between the agents for purposes of message exchange and subsequent event coordination. The communication between the agent container and the agent platforms is secured through the use of a largely available Secure Socket Layer technology (Andreadis, Klazoglou, Niotaki, & Bouzakis, 2014). 7.0 Negotiation The negotiation process involves communication between one agent and the other in a bid to acquire an operational plan. The negotiation process involves a two tier process between the product and resource. The first step or level involves task announcement and its corresponding qualification criteria. If the agents respond with an affirmative, the reply acts a BID for the task after which the product does an appraisal of the BID in order to come up with the best suited agent. The selected product source then reverts with an acceptance message for negotiation completion. Due to preoccupancy, if the best suited agent is not available, the system chooses on the next available one (Jana, Bairagi, Paul, Sarkar, & Saha, 2013). 8.0 Agent Architecture Multi-agent systems architecture is mainly classified according to the coordination system that is applied. The main coordination systems as learned from the coordination section 7.0 are either centralized or decentralized. A centralized system involves the collection of partial plans from each and every agent with the main intention of execution for final goal execution and delivery as a means of avoiding conflicts. On the other side, decentralized multi-agent coordination engages agent control through concerted communication with each other in order to come up with solutions and plans for upcoming arguments. The basis of architecture has been based on the mode of interaction since these characteristics identified schemes well. This makes it easy for categorization purposes. Multi-agent architecture mainly consists of mediator and facilitator agents which form a clearly centralized system. Classification done on agent function gives a wrong impression and may turn out to be ambiguous since engineering stages to be handled are multileveled. It is also safe for multi-agent architecture to identify architecture in terms of communication agents since it has now become a very important aspect in multi-agent systems development. The interaction mode indicates the potential problems that may arise from the suggested communication medium and resolution (Andreadis, Klazoglou, Niotaki, & Bouzakis, 2014). 8.1 Centralised Multi-Agent Architecture The centralised multi-agent architecture involves a system where a single agent manages other autonomous agents. Therefore, each agent has to get back to the coordinator agent for further relaying of information to the other agents that may have to contribute to a given task. Agent communication is handled by the central coordinator which happens to be the mediator for purposes of management and cooperation within the system. A facilitation program coordinates the agents through a reliable communication network that is capable of routing messages between the agents for regulation purposes. In centralized architectures, it is therefore apparent that the translation is offered by the coordinating agents to avoid mixed signals from being communicated in among agents. Mediator agents are also part of this robust system. These are entrusted to taking low level crucial decisions that affect the system both negatively and positively. Just like it happens in real life, the mediator agents are meant for intervening purposes whenever situations are getting out of hand. The mediator agents offer decisions when the system hits a snag for purposes of releasing the pressure that may otherwise have caused a backlog of commands to be executed. The mechanism used by this architecture thereby entrusts the brokerage power to the mediator agents interpret complex messages and task breakdown for ease of execution (Andreadis, Klazoglou, Niotaki, & Bouzakis, 2014). Figure 6: Centralised multi-agent systems (Andreadis, Klazoglou, Niotaki, & Bouzakis, 2014). The architecture of centralized multi-agent systems involves divergent agents seeking channels for communication and routing of information to other agents whose activity and execution patterns are monitored from a central agent. This form of coordination has been found to be very workable due to the existence of agents such as mediators and coordinators that are meant for standardization of knowledge and language. Sequence systems are concentrated in several agents which majorly seek to communicate thus bringing out this as the major characteristic of the centralized multi-agents interaction. The initial systems are expected to deal with information technology and its applications in order to build a workable virtual environment of existence say manufacturing. The determinative agent in the virtual systems is meant to manage the entire virtual environment and coordinate other agents. In the case of a manufacturing environment, these agents may include feature, design and geometric based on technology that is being deployed at the time of system enactment. The process plans are obtained from the from the machine agents which cluster the sequence through special pooling of tasks. Information is then obtained and updated seamlessly in order to ensure that it is converted into the much needed knowledge that eventually becomes part of the system (Andreadis, Klazoglou, Niotaki, & Bouzakis, 2014). Figure 7: Centralised multi-agent architecture (Andreadis, Klazoglou, Niotaki, & Bouzakis, 2014). The optimisation process of a centralised architecture system is designed through use of the A-design technology that is responsible of automation and design synthesis. This system is thus hierarchical by nature since it involves multi-levelled agents as shown in the figure above. The manager agents are required to make and modify designs that are created by design agents, give feedback for rectification if any and forward for execution. The maker agents are entrusted to the task of enhancing the configuration between fragment agents and instantiation-agents. Modification agents are on the other hand active when it comes to evaluation and improvement phase (Andreadis, Klazoglou, Niotaki, & Bouzakis, 2014). 8.2 Decentralised Multi-agent architecture The autonomity of the agents is respected within this system owing to the need to control the input externally through software of a person. These agents however interact with other agents in the system since they are bestowed with knowledge to plan on their own with regard to a common goal to be achieved by the system. This architecture however involves the preliminary decision plans that are associated with manufacturing, distribution and physical scheduling. Conventional planning mechanisms are cut down through the task decomposition within the multi-agent system without the need for a central control mechanism to do the same. These systems work with agents committed to making decisions and performing tasks within the manufacturing system for example with communication problems due to lack of a central agent dedicated to coordination process. This system is considered as an important approach towards industrial distributed systems in which operations are decentralised. This works effectively in environments such as databases, business networking and knowledge bases. Therefore an agent in the case of a decentralized multi-agent system is just but a separate computing entity that facilitates execution of a goal through independent control. The communication from one agent to the other is facilitated via parallel distributed messaging which coincidentally has to be asynchronous (Wong, Leung, Mak, & Fung, 2006). The system is designed to consist of agents that are dedicated to taking care of the functional details. The decentralized multi-agent systems are process oriented due to the need to execute short-term goals that eventually pile up to form one complete goal. The architectural outlook of an ordinary decentralised system is likely to consist of resource selection, detailed process planning and process selection agents which are supported by knowledge based, web and XML agents. Where knowledge based, web and XML agents are meant to act as the administration agents for interaction purposes. The web agent for example gives the agents a leeway to interact with the agents from their intended beginning to stop. The XML agents provide data base access to the agents via information displayed through web browsers and other tools as they arise. Knowledge base agents are known to handle the mathematical functions that need execution from time to time (Andreadis, Klazoglou, Niotaki, & Bouzakis, 2014). Software agents are incorporated within a decentralized system in order to offer a flow of information from one level to the other. An example is a company that requires a quick flow of information from one level to the other e.g. a supply chain company. The software agent is enabled to integrate the manufacturing framework whenever a supply chain is being developed. The manufacturing agent has to combine database and management in order to secure workable processes. The company and service provider agents are incorporated within these systems as a means of scheduling for manufacturing processes. The company agents are entrusted to the control functions that are associated with the abnormal events such as late materials procurement. The processes optimization agent forms a control agent within the general user interface in order to access the multi-agent system (Braga, Rossetti, Reis, & Oliveira, 2009). 9.0 Multi-agent System Communication Communication is deployed by agents when it comes to devising workable system solutions. At some point it is important that agents interact in a bid to maximise the internal functions that are associated with the network stability and utility. Coordination mechanisms that are considered as effective are constituted of regulatory agents whose interaction is only circumstantial and only when need arises. Some of the communication protocols that are involved in coming up with a well coordinated multi-agent system include; contract net protocol, market-based coordination, partial global planning, ACL, KQML and BDI-based negotiation protocol. Below are some of the descriptions of these protocols in order to offer a better understanding of how agents communicate in a multi-agent system (Lee & Kim, 2008). To begin with, the contract net protocol engages in three distinct stages of intercommunication namely announcement, bid and award messages. This protocol can only work in environments with well distributed and described scopes of functions to execute. Another form of communication identified is the partial global planning in which the agents communicate with each other via a local plan that is aimed at a global one. Market based coordination is a market based communication protocol that is deemed as competitive due to the auction design through which it operates. This system is best suited for a decentralized system as the protocol’s assumption is that the major decisions are localised. The KQML protocol is based on a high level language such as speech as a means of regulating the runtime between two or more agents. This protocol is mainly comprised of communication, message and content layers. Another protocol worth noting for this exercise is the ACL protocol which is founded on intelligence of physical agents. This protocol is very similar to KQML and FIPA when it comes to the prevailing semantics and syntax rules. Lastly the BDI-based negotiation protocol allows for agents to negotiate for tasks in an agreed goal system that withhold commitments in order to avoid conflicts from arising time and again (Lee & Kim, 2008). 10.0 conclusion Multi-agent systems are very important in dynamic scheduling of flexible manufacturing systems as a means of energy consumption reduction. This paper successfully characterises the multi-agent system for this purpose. Both architectures of multi-agent systems are covered to unearth the improvements that they may poses to the existing manufacturing system. System coordination or cooperation is established as one of the most vital step towards achievement of efficiency. A workable multi-agent system also requires cognition, interaction, organisation and work standardisation. 11.0 List of References Andreadis, G., Klazoglou, P., Niotaki, K., & Bouzakis, K.-D. (2014). Classification and Review of Multi-Agents Systems in the Manufacturing Section. Procedia Engineering , 69, 282-290. Badr, I. (2008). An Agent-Based Scheduling Framework for Flexible Manufacturing Systems. World Academy of Science, Engineering and Technology , 40, 363-369. Baffo, I., Confessore, G., & Stecca, G. (2013). A decentralized model for flow shop production with flexible transportation system. Journal of Manufacturing Systems , 32, 68-77. Braga, R. A., Rossetti, R. J., Reis, L. P., & Oliveira, E. C. (2009). Applying multi-agent systems to simulate dynamic control in flexible manufacturing scenarios. Artificial Intelligence and Computer Science Laboratory , 1-6. Caridi, M., & Cavalieri, S. (2004). Multi-agent systems in production planning and control: an overview. Production Planning & Control: The Management of Operations , 15 (2), 106-118. Cetnarowicz, K., & Kozlak, J. (2001). Multi-Agent System for Flexible Manufacturing Systems Management. 2nd International Workshop of Central and Eastern Europe on Multi-Agent Systems, (pp. 51-60). Krakow. Chen, K.-Y., & Chen, C.-J. (2010). Applying multi-agent technique in multi-section flexible manufacturing system. Expert Systems with Applications , 37, 7310–7318. Erol, R., Sahin, C., Baykasoglu, A., & Kaplanoglu, V. (2012). A multi-agent based approach to dynamic scheduling of machines and automated guided vehicles in manufacturing systems. Applied sof computing , 12, 1720-1732. Ferber, J. (1993). Modele de systemes multi-agents-Du reactif au cognitif. Proceedings of INFAUTOM'93 `du traitement reÂparti aux systeÁmes multi-agents et aÁ l'autonomie des systemes, (pp. 26-56). Toulouse. Guo, Q., & Zhang, M. (2000). A novel approach for multi-agent-based Intelligent Manufacturing System. Information Sciences , 179, 3079-3090. Guo, Q.-l., & Zhang, M. (2010). An agent-oriented approach to resolve scheduling optimization in intelligent manufacturing. Robotics and Computer-Integrated Manufacturing , 26, 39-45. He, N., Zhang, D. Z., & Li, Q. (2014). Agent-based hierarchical production planning and scheduling in make-to-order manufacturing system. Int. J. Production Economics , 149, 117-130. Jana, T. K., Bairagi, B., Paul, S., Sarkar, B., & Saha, J. (2013). Dynamic schedule execution in an agent based holonic manufacturing system. Journal of Manufacturing Systems , 32, 801-816. Kouiss, K., Pierreval, H., & Mebarki, N. (1997). Using multi-agent architecture in FMS for dynamic scheduling. Journal of Intelligent Manufacturing , 8, 41-47. Lee, J. H., & Kim, C. O. (2008). Multi-agent systems applications in manufacturing systems and supply chain management: a review paper. International Journal of Production Research , 1 (1), 233–265. Leitao, P. (2009). Agent-based distributed manufacturing control: A state-of-the-art survey. Engineering Applications of Artificial Intelligence , 22, 979-991. Li, X., Zhang, C., Gao, L., Li, W., & Shao, X. (2010). An agent-based approach for integrated process planning and scheduling. Expert Systems with Applications , 37, 1256-1264. Lim, M. K., & Zhang, Z. (2003). A multi-agent based manufacturing control strategy for responsive manufacturing. Journal of Materials Processing Technology , 139, 379-384. Lim, M. K., & Zhang, Z. (2012). A multi-agent system using iterative bidding mechanism to enhance manufacturing agility. Expert Systems with Applications , 39, 8259-8273. Lim, M. K., Zhang, Z., & Goh, W. T. (2009). An iterative agent bidding mechanism for responsive manufacturing. Engineering Applications of Artificial Intelligence , 22, 1068–1079. Lin, J., & Long, Q. (2011). Development of a multi-agent-based distributed simulation platform for semiconductor manufacturing. Expert Systems with Applications , 38, 5231-5239. Monostori, L., Vancza, J., & Kumara, S. T. (2006). Agent-Based Systems for Manufacturing. Annals of the CIRP , 55 (2), 697-720. Nie, L., Bai, Y., Wang, X., Liu, K., & Cai, C. (2012). An Agent-based Dynamic Scheduling Approach for Flexible Manufacturing Systems. Proceedings of the 2012 IEEE 16th International Conference on Computer Supported Cooperative Work in Design, (pp. 59-63). Shanghai. Ouelhadj, D., Hanachi, C., & Bouzouia, B. (2000). Multi-agent architecture for ditributed monitoring in flexible manufacturing systems (FMS). Proceedings of the 2000 IEEE International Conference on Robotics & Automation, (pp. 2416-2421). San Francisco, CA. S., B., & K., S. (2001). An Agent-based Approach to the Control of Flexible Production Systems. 8th IEEE Int. Conf. on Emerging Technologies and Factory Automation (ETFA 2001), (pp. 168-174). Antibes Juan-les-pins. Shen, W. (2002). Distributed Manufacturing Scheduling Using Intelligent Agents. IEEE intelligent systems , 86-94. Shen, W., Hao, Q., Yoon, H. J., & Norrie, D. H. (2006). Applications of agent-based systems in intelligent manufacturing: An updated review. Advanced Engineering Informatics , 20, 415-431. Shen, W., Wang, L., & Hao, Q. (2006). Agent-Based Distributed Manufacturing Process Planning and Scheduling: A State-of-the-Art Survey. IEEE transactions on systems, man, and cybernetic - part c: Applications and reviews , 36 (4), 563-577. Sudo, Y., & Matsuda, M. (2013). Agent based manufacturing simulation for efficient assembly operations. Forty Sixth CIRP Conference on Manufacturing Systems 2013, (pp. 437-442). Kanagawa. Wang, L.-C., & Lin, s.-K. (2009). A multi-agent based agile manufacturing planning and control system. Computers & Industrial Engineering , 57, 620-640. Wong, T. N., Leung, C. W., Mak, K. L., & Fung, R. Y. (2006). Dynamic shopfloor scheduling in multi-agent manufacturing systems. Expert Systems with Applications , 31, 486-494. Xiang, W., & Lee, H. P. (2008). Ant colony intelligence in multi-agent dynamic manufacturing scheduling. Engineering Applications of Artificial Intelligence , 21, 73-85. Read More

Some of the identifiable features of an agent include the capability to represent the environment partially or by perception, ability to communicate with other agents, ability to produce son agents and the autonomy behaviour in which the objectives are preserved. An agent is split into three layers namely the static knowledge layer, expertise layer and the communication layer. A static knowledge layer as applied in the structuring of agents is said to contain knowledge within itself as a host to the specific memory that is meant to remember every single step that concerns its environment of existence.

The expertise layer usually contains the treatment representation for a given knowledge action that is supposed to perform the intended action within a given time frame through logical expressions or algorithms guided by a certain set of irrefutable production rules. The communication layer identities all the communication tools that build up a successful communication protocol between one agent and the other. Depending on the form of communication that is set in place, this layer takes into account of all messages that are sent and received (Kouiss, Pierreval, & Mebarki, 1997).

Figure 1: Structure of an agent (Kouiss, Pierreval, & Mebarki, 1997). 3.0 What's multi-agent system? Based on the understanding of an agent, a multi-agent system is an artificial system which is made up of diverse autonomous agents which are well collaborated in order to achieve a major organisational goal. These agents usually work simultaneously while pursuing individual goals that contribute to the achievement of a single final objective (Kouiss et al., 1997). Each independent agent involved in a multisystem is perceived in terms of its usefulness as a problem solver within a given environment due to the level of artificial intelligence that is bestowed upon it.

The feasibility of an agent as an optimal problem solver lies entirely on the heterogeneity, hierarchy and structure as shown in the figure below (Lee & Kim, 2008). Figure 2: Multi-agent system layout (Nie, Bai, Wang, Liu, & Cai, 2012). Agents possess problem solving capability and are therefore meant to interact with each other via a given communication protocol in order to cooperate. Multi-agent systems interactions are further categorised into cooperative and self-interested interactions which depict the optimal course of action that a given solution takes.

The synchronous and asynchronous communication protocols that are involved are meant to boost the accuracy of information relayed from one level to the other to allow for global planning. Comparing these complex interactions to real world situations, it shall be noted that the overriding resolution setups work in the same way of deployment (Andreadis, Klazoglou, Niotaki, & Bouzakis, 2014). The adaptability concept of multi-agent systems refers to the changes that have been monitored over time through trial and error in order to come up with a workable system.

The proactive nature of taking initiatives by these systems exhibit a goal oriented comportment. Multi-agent systems thus act out of reasoning to adapt with their interaction patterns and hence end up being characterised by such notions as beliefs, goals and plans that develop over time. The optimal courses that are acted upon by agents making up a system are based in the preferences or goals that are set for the entire system. This means that the autonomy of a given agent may be affected by the rational choices that may be undertaken by another agent that runs concurrently in order to deliver a given goal. 4.0 Characteristics of Multi-agent systems Multi-agent characteristics are dependent on the basic properties attached to agents.

Therefore, in order to create an understanding on characteristics of multi-agents it is mandatory to understand the basic properties of agents. To begin with, characteristics of agents are observed with regard to the environment, belief and knowledge, environmental state and actions related to environmental change.

Read More