Agent-Based Modeling in Architecture - Term Paper Example

Agent Based Modeling in Architecture Table of Contents Introduction 2 Agent-based modeling 3 Agents 4 Concepts of agent and agency 5 Emergence 6 Agents and the environment 8 Programmable agents 10 Types of agents 10 Simple reflex agents 11 Model based reflex agents 12 Goal-based agents 12 Utility-based agents 12 Learning agents 13 Properties and behavior of agent-based models 13 Learning 14 Coordination in multi-agent systems 15 Applications of Multi-Agent Research 16 Conclusion 17 References 18 Introduction The research and development in the field of the artificial intelligence has offered lot of new technology infrastructures and paradigms. The implementation of the modern and up-to-date techniques of artificial intelligence has offered huge advantages regarding machine learning, robotics, and automation of a lot of operating behaviors. This paper presents a detailed analysis and investigation of the Agent-based modeling and multi-agent systems. This paper will present a deep insight into the development, usage, technology infrastructure, type and operating of these Agent-based modeling. Walsh (2009) states that various researchers and scholars have carried out research on the efficiency of agent based modeling, and they concluded that agent based modeling produces more data and information as compared to the equations based modeling (Walsh, 2009). According to Macal & North (2006), the Agent Based Modeling is acknowledged or identified with different names like that agent-based modeling (ABM), agent based systems (ABS), and individual based modeling (IBM). These are extensively and commonly utilized identification names of the Agent Based Modeling (Macal & North, 2006). According to Walsh (2009), the agent-based modeling is becoming more and more widespread. The main reason behind this aspect is the increasingly complex world operation those are requiring more automation of operations. In addition, the customary modeling technologies are no longer as appropriate as they were in the past. An instance of old technology implementation is the area that is deregulation of the electric power industry. In case of modeling regarding economic marketplaces those are conventionally dependent upon the ideas of homogeneous agents, ideal marketplaces, and long-run equilibrium for the reason that these suppositions are established on the computationally tractable and problems analytically. Furthermore, we are becoming capable to take an additional sensible view of these economic systems through ABM. However, there is another important reason of agent-based modeling; the data is arranged into databases at more improved levels of granularity. In addition, at the present, we have more micro-data that is helpful in managing the business operations and activities in an appropriate manner (Walsh, 2009). Bonabeau (2002) outlined that in agent based modeling (ABM), a system is developed through a combination and collection of self governing and autonomous decision making entities those are acknowledged as agents (will be discussed in coming sections). Every autonomous agent makes the decisions on the basis of the provided set of rules (Bonabeau, 2002). Agent-based modeling According to Bonabeau, Dorigo, & Theraulaz (1999), the Agent-Based Modeling (ABM) is a relatively modern computational modeling model that is based on the modeling of phenomenon like dynamical structures of cooperating intelligent agents (Bonabeau et al., 1999). In addition, the Agent-Based Modeling is not a technology but it is an approach or technique. The Agent Based Modeling frame of work composed of an arrangement of its components/ units or agents. However, according to various researchers, an alternative to Agent-Based Modeling is conventional discrepancy equation modeling. But, this is not true, like a group of differential equations, in which every equation demonstrating the dynamics of one of the system's units that is similar to the agent-based model. As the Agent-Based Modeling approach is opening towards the major recognition (Bonabeau et al., 1999). It is necessary to discuss basic concepts before going into the depth of Agent-based modeling. Agents Macal & North (2006) outlined that the term “agent” has different implications. However, there are no common definitions for the term “agent”. A number of modelers believe that some kind of autonomous component like model, software, individual, etc. could be considered as an agent. However, a self-governing component’s activities are ranged from complex adaptive AI to primitive reactive decision rules. According to some other researchers a component’s behavior could also be taken as agent (Macal & North, 2006). According to Bonabeau, Dorigo, & Theraulaz (1999), an agent in the Agent-Based Modeling can perform a variety of behaviors that are suitable for the system. In addition, the Agent-Based Modeling can be consumed, produced, or applied to carry out a particular job. Furthermore, recurring competitive connections among agents are a characteristic of agent-based modeling that depends upon the control of agents/computers to investigate dynamics out of the accomplishment of pure mathematical techniques. In simple words, an agent-based model is developed using a structure of agents as well as the associations among them. However, still a simple agent-based model demonstrates complex performance models as well as offer helpful information regarding the dynamics of the real-world structure that it followed. Furthermore, agents could be able to develop, demonstrate unexpected behaviors to emerge. Sophisticated Agent-Based Modeling occasionally includes evolutionary algorithms, neural networks, or additional learning methods to permit practical learning as well as adaptation (Bonabeau et al., 1999). Concepts of agent and agency According to Macal & North (2006) an agent comprises two sets of rules. The first type of rule that an agent contains is base-level rules intended for behavior and a higher-level set of rules. The higher sets of rules are intended to transform the rules. In addition, in an agent, base level rules offer responses to the environment however, the higher level rules those change existing rules offer “adaptation” (Macal & North, 2006). Macal & North (2006) outlined that computer science aspect of agency is highlighting the vital feature of independent behavior. The basic characteristic of an agent is the ability to formulate self-governing judgments. This necessitates agents to be dynamic rather than just passive (Macal & North, 2006). As Bonabeau, Dorigo, & Theraulaz, (1999), outline some of the main characteristics of an agent those are given below (Bonabeau et al., 1999): An agent is a special, a distinct individual holding some rules and characteristics those are leading to its decision making and behaviors ability. Agents are self-reliant. Agents have discreteness requirement, in other words, an agent has a limit as well as one can simply conclude whether something is division of an agent or not, or is a common feature. An agent is positioned and operates in an environment by means of its connection along with other linked agents. Each agent possesses some kind of protocols for communication with other agents, like that for communication. This protocol is also used to respond to the environment. Agents also possess the capability to recognize as well as differentiate the characteristics of other communicating agents. An agent can be goal directed, in other words, an agent contains some kind of goals those are achieved with respect to its behaviors. This permits an agent to evaluate the result of its behavior comparative to its objectives. Another main aspect of an agent is nature of self-directing and autonomous functioning. For instance, an agent is capable to work autonomously in its environment as well as in its communication with other agents, at any rate over a partial variety of states that are of interest. Flexibility is also a main aspect of an agent. An agent possess the capability to adapt and learn its behaviors those are foundational on experience. This necessitates a number of shapes of memory. An agent also contains rules that can transform its policy of behavior. Emergence The paradigm of agent-based modeling has initiated in late 1940s. Since from that time this area of research has taken lot of huge development steps. However the huge development in these areas was seen after 1990s. The main reason behind this extensive evolution is the widespread computer based system regarding the technology implementation (Bonabeau et al., 1999). However, Russell & Norvig (2003) outline that the development or emergence of agent-based modeling paradigm has its roots in Von Neumann machine. Therefore, a lot of new developments have emerged. However, one of the earliest developments in the area of agent-based modeling was done by the Thomas Schelling's segregation model. This model was developed in 1971. In this model Thomas Schelling presented the basic concept with the help of coins and graph paper rather than computers. In addition, Thomas Schelling was the first person who outlined the basics of agent-based models where independent and self governing agents are interacting to communicate and carry out some definite operations. These individuals’ goals are collected or summed up to produce the final output (Russell & Norvig, 2003). Also, another main development in the agent-based modeling paradigm was done in early 1980s. In this scenario Robert Axelrod has hosted a game of Prisoner's Dilemma approaches as well as had them working together in an agent-based way to decide a winner. In addition, Robert Axelrod also developed some of other agent-based models in the area of political science. These agents are developed to examine events as of ethnocentrism in the direction of the dissemination of culture (Walsh, 2009). Another development was done by Craig Reynolds in 1980s, when he developed the initial biological agent-based model that was holding the social characteristics. He attempted to model the actuality of dynamic biological agents, recognized like artificial life, a term invented by Christopher Langton (Bonabeau et al., 1999). The concept of the agent-based modeling all through the mid-1990s spotlighted over the aspects of developing and designing effective teams and offering effective communication that is needed for the organizational effectiveness, as well as the performance of social networks (Bonabeau et al., 1999).However, with the emergence of StarLogo, SWARM, NetLogo and RePast, a huge progress has been done in the field of agent-based modeling (Bonabeau et al., 1999). More recently Ron Sun developed the techniques intended for foundation agent-based simulation on forming of human cognition, recognized like cognitive social simulation, this was developed in the 2006 (Castiglione, 2006). Suzanne Lohmann, Bill McKelvey, Dwight Read, Dario Nardi, and others at UCLA have as well developed important assistance in corporate behavior as well as decision-making (Macal & North, 2006). Agents and the environment According to Odell, Parunak, Fleischer, & Brueckner (2002), “an environment offers the circumstances under which an agent (object or entity) operates and exists. In addition, without an operating environment that is held by an agent could not be able to work effectively. Also, an agent is not able to act or sense. An environment offers the circumstances underneath an agent/entity/object to exist and operate. It describes the characteristic of the world in that an agent will operate. Therefore, the effectively designing of agents needs cautious concerns of equally the communicational and physical features of their operating circumstances and environment (Odell et al., 2002). Furthermore, (Odell et al., 2002) outline two main features of an understanding environment and agent interaction: Each agent has an operating environment. This environment does not depend on the what is the agent’s architecture or philosophy. Being responsive of the agent’s environment facilitates its designer to obtain stronger interaction by means of architecture-dependent ways. According to (Odell et al., 2002), an entity is identified as an agent if it has some level of independence, that is, if it is noticeable from its operating environment through a number of types of temporal, spatial or functional characteristics. That is, an agent has to be specialized (Odell et al., 2002). In addition, an agent could control the equally communication and physical environment intended for its designated place foundational on the processing and global physical requirements. For instance, Figure 1 demonstrates having one environment agent linked with the environment. This environment agent is as well linked with the population of domain agents it possesses (Odell et al., 2002). Figure 1- Environment and Agent Image Source (Odell et al., 2002) or [http://www.jamesodell.com/Agents_and_environment.pdf] Programmable agents As Al-Shaer (1999) has stated that intelligent software agents are offering a lot of valuable facilitates as these agents are handled through the malleable and programmable tools those reduce users’ control. The automation of the agents is a really good thing however we need to define some control over tasks essential for daily life. In this way we could be able to mold the reaction of the agents according to our desires and needs (Al-Shaer, 1999). As Nardi, Miller, & Wright (1998) stated that programmable, collaborative agents facilitate users to obtain additional value out of the data they are previously employing, and for jobs they are previously carrying out. In addition, by means of programmable agents we have to put less effort and extensive level of objectives specification. Furthermore, the programmable agents will as well offer novel functionality because people determine new applications for them. Nardi, Miller, & Wright (1998) outline the application of the programmable agents through development of “Apple Data Detectors” that is developed to hold up an evolutionary procedure through which users outline their own tools to the maximum degree as probable (Al-Shaer, 1999) & (Nardi et al., 1998). Types of agents According to Castiglione (2009), the intelligent agents are based on the artificial intelligence paradigm and they are strongly associated to agents in financial side. In addition, they are the versions of the smart agent models those are investigated in ethics, cognitive science and the philosophy of practical reason. Furthermore, a lot of interdisciplinary socio-cognitive representation and computer social simulations are also linked with these intelligent agents (Castiglione, 2006). However, intelligent agents are as well directly linked to software agents (an independent software application that performs jobs on behalf of clients). In computer science, the expression intelligent agent can be employed to refer to a software agent that has a number of intellects, in spite of if it is not a balanced agent by Russell and Norvig's description of agents. For instance, autonomous applications employed for operator help or data mining (occasionally represented as bots) are as well acknowledged as "intelligent agents" (Castiglione, 2006). Russell & Norvig (2003) has offered the clear distinction among the intelligent agents. They have categorized the intelligent agent into 5 categories. This division is based on the extent of apparent intelligence as well as potential of these agents. The further description is given below: (Russell & Norvig, 2003): Simple reflex agents Simple reflex agents perform actions simply on the foundations of the present percept. The agent operations are foundational upon the condition and action rule. It can be presented as: if condition A is true then take some action. This simple reflex agent operates simply when the environment is completely observable. A number of reflex agents are able to as well hold information on their present states those permit them to ignore circumstances whose actuators are previously activated (Russell & Norvig, 2003). Model based reflex agents Model-based agents are capable to manage and control partially visible situations. Its present state is stored within the agent that is upholding a number of jobs. In addition, they also explain the fraction of the world that is not able to be seen. This performance necessitates data and information on how the world performs or operates. This extra information is inclusive to the “World View” of agent model. A model-based reaction agent continues to track the present state of the world by means of an interior model. It then selects an act in the similar way as the reaction agent (Russell & Norvig, 2003). Goal-based agents Goal-based agents are also similar to model based agents those are holding the information about situations that are wanted. This permits the agent a method to decide among numerous situations/conditions and potential, choosing the one that arrives at a objective state (Russell & Norvig, 2003). Utility-based agents Goal based agents simply differentiate among objective states as well as non objective states. It is probable to describe determinations of how attractive a particular condition is. However, these calculations can be acquired by making use of a utility function that maps a condition to calculate the utility of the condition (Russell & Norvig, 2003). Learning agents Learning has a benefit that it permits the agents to primarily function in unidentified situations as well as to become more capable than its preliminary knowledge is only permitted (Russell & Norvig, 2003). Properties and behavior of agent-based models Srbljinovic & Skunca (2003) outline some of the main properties and behavior of agent-based modeling, as agent-based models are attractive and appealing that are outcomes of the combined level (that frequently neither apparent, nor expectable) processing, in a lot of cases when the supposition on particular agent properties those are extremely simple. Specifically, the potential of collecting complex as well as intriguing properties of agent happen not a great deal from the in-built regulations of individual agent actions, as of the complication of the network of communications between the agents. Furthermore, this massive amount of agents, and the huge number of complexity of their communications are the fundamental cause why in the majority cases the formal mathematical reasoning of outcomes of an agent-based model is not probable (Srbljinovic & Skunca, 2003). In this scenario Axelrod & Axelrod (1997) state that it is as well the cause why the matters of complexity remained comparatively under-explored in anticipation of recent established transformations in technology. Specifically, as scientists frequently make a decision to pay concentration to “problems explanation through the theoretical and influential methods previously at hand”. These are pushed to the edge of technical investigation, either for the reason that they are considered not to be applicable, or since their investigating would require unavailable method. Therefore, simply after fresh advancements in the expansion of computational technology it has been facilitated huge simulation experiments, the subjects of emergent complication approached closer to the heart of scientific research (Axelrod & Axelrod, 1997). Also the above stated modeling of bottom-up outcomes i.e. are the effects producing at the individual level as well as controlling the shared one. Additional composite agent-based models are as well able of modeling top-down outcomes, happening at the combined level as well as controlling the level of individual agents. Learning Jafari (2002) has outlined that agent-based modeling learning aspect indicates that the occurrence of intelligent agents is valuable. In addition, developing new human-like structures by means of intelligent agents builds users’ communications with the computer in a very smoother way. Furthermore, the research has shown that academic agents can facilitate learners to build up an emotional connection with the agent, and helping in their pleasure of the learning circumstances. Besides this, the learner’s building of a social association by a pedagogical agent is a main method in development of communication as well as promoting learning inside a computer-based learning structure (Jafari, 2002). According to Chang, Ho, & Kaelbling (2004) agent-based modeling is a novel method of performing science that has built and transformed the concepts as well as methods of complexity theory. It engages the investigation of a lot of actors and their communications. The models initiated by straightforward rules of learning and postulations however will exhibit complex behaviors. This tool is well-matched by qualitative and quantitative research techniques (Chang et al., 2004). . Coordination in multi-agent systems According to Kok (2006), in the analysis of multi-agent systems we pay attention to completely cooperative multi-agent systems since the entire agents contribute to a common objective. Every agent chooses actions independently, however it is the consequential in joint action that creates the results. In addition, a main feature in similar structures is consequently the difficulty of synchronization: the procedure that makes sure that the individual judgments of the agent’s outcomes in optimal judgments for the group like a whole. Furthermore, without recognizing the option of the other agent, the objective of the agents is to choose an accomplishment that would be different from the other agents. The coordination among agents is a problem that addresses the significant question how the agents should choose their individual proceedings (Kok, 2006). In this scenario, Cebulla (2005) outlines that it is supposed that agents identify the results of every potential joint acts, and therefore are capable to decide the set of the entire most favorable cooperative actions, this thing is called equilibrium in this situation. At the present the coordination difficulty is making simpler to the problem of agreeing on a particular cooperative action from this set (Cebulla, 2005). In addition, Cebulla (2005) presented three diverse solution methods to this problem: social conventions, communication, and learning. Communication permits every agent, in a predefined succession, to notify the further agents of its act, limiting the option of the other agents. While, social conventions are restraints on the accomplishment of alternatives of the agents. Earlier the agents consent upon a precedence ordering of agents as well as actions. When a procedure has to be chosen, every agent gets actions that the agents want to perform with a higher priority, after that chooses its own action consequently. A widespread instance of a social convention is the right-of-way law in traffic in through which an agent approaching from the right has precedence in passage a crossroad. The agent approaching from the right recognizes it has the uppermost priority as well as decides to drive in the course of (action by highest priority). Furthermore, the agent is able to derive the action of the initial agent, as well as decides to stop. At the end by learning techniques could be implemented to find out the behavior of the agents in the course of frequent interaction (Cebulla, 2005). Applications of Multi-Agent Research Bernon, Chevrier, Hilaire, & Marrow (2005) stated that a variety of techniques for stimulating self association inside multi-agent structures means that Multi-Agent system have the possible to support to a range of applications. In Multi-Agent system it is illustrate that upon the feature of self-organization to formulate the implementation in a more efficient way. Additionally of Multi-Agent systems applications are considered in the regions of flood forecasting, timetabling, localization, land use, allocation and tracking. In addition, we have its better implementation in meshing for the wireless network and traffic simulation (Bernon et al., 2005). Below I will mention some of the main implementation areas of the Multi-Agent systems those are offered by the Bernon, Chevrier, Hilaire, & Marrow (2005): (Bernon et al., 2005) Aircraft maintenance Wireless communications and collaboration Electronic book buying coalitions Military data mining Joint mission planning Supply-chain management Financial portfolio management Military logistics planning Multi-agent system for web searching A multi-agent structure intended for intelligent observing of airline operations Natural language text mining and processing Agent-based electrical energy e-market Legged robotic surveillance ”smart guardian” Conclusion In this research I have presented a detailed analysis of the agent-based modeling. It is analyzed that intelligent agent is a group of self-governing software tools connected with other through some databases and applications execution inside one or several computer associated atmosphere. The main purpose of an intelligent agent is to facilitate a client to use a system in a better way, and to manage as well as interact with a computer application. Agent-based technology structures are developed through the artificial intelligence (AI) and comprise a degree of independent problem-solving capability. This research has tried to offer a deep insight into to the overall analysis of the agent-based modeling. I hope this will provide a comprehensive analysis of the agent-based modeling. References Achorn, E., 2005. Integrating Agent-Based Models with Quantitative and Qualitative Research Methods. [Online] Available at: HYPERLINK "http://74.125.153.132/search?q=cache:55LX-xylVnMJ:www.aare.edu.au/04pap/ach04769.pdf+learning+Agent-based+modeling&cd=6&hl=en&ct=clnk&gl=pk" [Accessed 21 January 2010]. Al-Shaer, E.S., 1999. Programmable Agents for Active Distributed Monitoring. Lecture Notes In Computer Science Proceedings of the 10th IFIP/IEEE International Workshop on Distributed Systems: Operations and Management: Active Technologies for Network and Service Management, 1700, pp.19 - 32. Axelrod, R. & Axelrod, R.M., 1997. The complexity of cooperation: agent-based models of competition and Coopertaion. New Jersy: Princeton University Press. Bernon, C., Chevrier, V., Hilaire, V. & Marrow, P., 2005. Applications of Self-Organising Multi-Agent Systems: An Initial Framework for Comparison. Informatica (Slovenia), 30(1), pp.73-82. Bonabeau, E., 2002. Agent-based modeling: Methods and techniques for simulating human systems. Proceedings of the National Academy of Sciences, 99(3), pp.7280-7287. Bonabeau, E., Dorigo, M. & Theraulaz, G., 1999. Swarm Intelligence. From Natural to Artificial Systems. 1st ed. New York: Oxford University Press. Castiglione, D.F., 2006. Agent based modeling. [Online] Available at: HYPERLINK "http://www.scholarpedia.org/article/Agent_based_modeling" [Accessed 22 January 2010]. Cebulla, M., 2005. Modeling Coordination in Multi-Agent Systems by Knowledge Diffusion. In CIMCA Proceedings of the International Conference on Computational Intelligence for Modelling, Control and Automation and International Conference on Intelligent Agents, Web Technologies and Internet Commerce Vol-2 (CIMCA-IAWTIC'06) - Volume 02. Vienna, 2005. IEEE Computer Society Washington, DC, USA. Chang, Y.-H., Ho, T. & Kaelbling, L.P., 2004. All learning is local: Multi-agent learning in global reward games. [Online] Available at: HYPERLINK "http://people.csail.mit.edu/ychang/nips03.pdf" [Accessed 21 January 2010]. Jafari, A., 2002. Conceptualizing Intelligent Agents for Teaching and Learning. Educause Quarterly, 3(1), pp.28-34. Kok, J.R., 2006. Coordination and Learning in Cooperative Multiagent Systems. [Online] Available at: HYPERLINK "http://dare.uva.nl/document/35328" [Accessed 23 January 2010]. Macal, C.M. & North, M.J., 2006. Tutorial on agent-based modeling and simulation part 2: how to model with agents. In Winter Simulation Conference Proceedings of the 38th conference on Winter simulation. Monterey, California, 2006. Winter Simulation Conference. Nardi, B.A., Miller, J.R. & Wright, D.J., 1998. Collaborative, programmable intelligent agents. Communications of the ACM, 41(3), pp.96 - 104. Odell, J., Parunak, H.V.D., Fleischer, M. & Breuckner, S., 2002. Modeling Agents and their Environment. Lecture Notes on Computer Science , 2585, pp.16-31. Russell, S.J. & Norvig, P., 2003. Artificial Intelligence: A Modern Approach Second Edition. 2nd ed. Upper Saddle RiverNJ US: Prentice Hall. Srbljinovic, A. & Skunca, O., 2003. An Introduction Agent Based Modelling and Simulation of Social Processes. Interdisciplinary Description of Complex Systems, 1(1-2), pp.1-8. Walsh, J., 2009. Artificial Intelligence in an Agent-Based Model. Thomas Jefferson High School for Science and Technology Burke, Virginia SOURCE: [http://docs.google.com/viewer?a=v&q=cache:1g0ZhFPWWsEJ:www.tjhsst.edu/~rlatimer/techlab09/WalshPaperQ4-09.pdf+ai+AND+Agent-based+modelling&hl=en&gl=pk&sig=AHIEtbQLRZDf3cqEKi], pp.1-8. Read More