Multi-Agent Systems--论文代写范文精选

仿真技术的早期发展,有模拟和理论方面,我们可以看到互联网流行的分布式人工智能和多主体系统。数学理论,如微分方程和随机过程,需要描述确定性和随机过程,以预测或描述现象如种群的动态。博弈论是一种应用数学，来理解经济行为。下面的essay代写范文会进行论述。

Abstract
  THErst chapter has discussed the issue of how cognitive science can contribute to social sciences in order to explain interactive social behaviour. In chapter 5, we will develop a new cognitive agent-based computational social simulation model (RBot) in an attempt to satisfy the need for new theories and models. In this chapter, we will discuss Multi-Agent Systems (MAS) as one of the theories for the development of such a new model. Multi-Agent Systems (MAS) is the name for a new approach of designing, analysing and implementing complex adaptive software systems (Jennings, Sycara, & Wooldridge, 1998). This approach has emerged out of a history of theory and applications in the area of (social) simulation models. 

  Figure 2.1 shows a selection of simulation theories and techniques that have been developed over the years and many of them have been applied in the eld of MAS and social simulations. The early development of simulation techniques at the left side of the gure are primarily based on mathematics. Moving to the right, there are simulations and theories that exist because of the introduction of the computer and at the bottom right, we can see the inuence of Internet on the upcoming popularity of Distributed AI and Multi-Agent Systems. 

  Mathematical theories, e.g. differential equations and stochastic processes, require specic skills for describing deterministic and stochastic processes in an attempt to predict or describe phenomena such as the dynamics of populations (Volterra, 1926). Game theory, discussed later on in this chapter, is a form of applied mathematics to understand economic behaviour based on interactions or games between rational agents who try to maximise their prots. Game theory still has great inuence in describing outcome strategies in simulations. In the 1960s, the rst computer simulation models were developed with help of queueing models, discrete-event simulations (Birthwistle, Dahl, Myhrhaug, & Nygard, 1973) and system dynamics (Forrester, 1991). Around the same time, micro-simulation or micro-analytical simulation models (MSMs) emerged as a simulation technique. 

  Another approach that emerged was cellular automata whose behaviour emerges from properties of local interactions. The basic model of a cellular automata is a two-dimensional grid of cellsthe structure of squared paperin which a cell reacts only with its neighbouring cells. The simulation uses a timestep generator and at every time-step, every cell reacts to its neighbours according to a set of rules. In this way a stepwise reaction will ow through the grid and inuences the overall behaviour of the system. 
To summarise, cellular automata model a world in which space is represented as a uniform grid, time advances by steps, and the `laws' of the world are represented by a uniform set of rules which compute each cell's state from its own previous state and those of its close neighbours. (Gilbert & Troitzsch, 1999, p. 122) In the 1980s, articial intelligence, a separate eld concerned with intelligence of the individual, became interested in the interaction and distribution of intelligence, known as Distributed Articial Intelligence (DAI). Ferber (1999, p. 24) gives a short description about the history of the roots of DAI that today has resulted into Multi-Agent Systems. 

  DAI started around the eighties with the model of the blackboard system (Erman, Hayes-Roth, Lesser, & Reddy, 1980); a model consisting of Knowledge Sources (KSs) organised in a star-topology with 18 in its core the `blackboard'a place where the KSs can share their knowledge and which is managed by a separate control device that has the task to prevent and coordinate conicts of access between these KSs. Around the same time another type of control system (cf. Lenat & Brown, 1984; Lenat, 1975) came up that solved problems with help of a community of specialists. Hewitt (1977) described control structures in which he did not consider processes as a sequence of choices, but he tended to think in terms of distributed systems considering control structures as patterns of message passing between active entities called actors (Ferber, 1999, p. 25) 1 . 

  The roots of DAI are the foundations of today's two approaches in DAI: Distributed Problem Solving (DPS) and Multi-Agent Systems (MAS) (Bond & Gasser, 1988). DPS takes a pure engineering approach to distributed systems, i.e. it is concerned with how to build functioning, automated, coordinated problem solvers for specic applications (Bond & Gasser, 1988, p. 4) (as cited by Van den Broek, 2001, p. 21). The approach taken by DPS is a top-down approach in which tasks are decomposed into smaller tasks appointed to specialists. 

  A DPS is. . . a top-down designed system, since agents are designed to conform to the problem-solving requirements specied at the top. Within this top-down task decomposition approach, the individual components are considered to be of secondary importance to the need of the overall system. The agents themselves have limited autonomy because their role in solving the overall problem is usually designed-in, with coordination rules included. (emphasis added: Van den Broek, 2001, p. 22) On the other hand, a MAS is constructed taking the bottom-up approach. The agent itself is now the centre of attention and not the system as a whole. In a MAS, the agent has more autonomy, and control is delegated to the agent, i.e. the overall outcome of solving the problem is not determined by a strictly topdown organised control system, as in the case of DPS, but by (social) interactions between independent and rational agents making decisions that satisfy their own (be it social) goals. 

  In other words, MAS can supply us with the means to study what the aspects of actors are that plausibly explain interactive (social) behaviour (see chapter 1; research question 1). Hence, MAS is concerned with the behaviour of a collection of distributed autonomous (heterogeneous) agents aiming at solving a given problem. The characteristics of a generic MAS are (Jennings et al., 1998): • 
Each agent has incomplete information, or capabilities for solving the problem, thus each agent has a limited viewpoint; • 
There is no global system control; • 
Data is decentralised; and • Computation is asynchronous Inferred from these characteristics, the main component studied in MultiAgent Systems is the autonomous agent and its associated behaviour in an environment. 

  The approach taken in this chapter is to explain MAS starting with the individual agent and subsequently going towards a system of (cognitive) social agents that considers relations, coordination and interaction between agents. The purpose of the chapter is not to provide a detailed review of the eld of MAS, but to accustom the reader with the paradigm MAS4 . This chapter is structured as follows. First, in section 2.1, the notion of agency is discussed to clarify the term agent. The term agent is applied in many research elds, e.g. biology, economics, cognitive science, social science and so on, and therefore has become somewhat blurred. In section 2.2, we approach the agent as a human-like entity in which components, e.g. perception, cognition, action and so on, form the basis for constructing an agent architecture. Section 2.3 describes two agent typologies: the cognitive agent commonly used in AI and the social agent. Whereas the cognitive approach focuses on the internal working of the agent, the social approach studies how (socially) well an agent is embedded in its social-cultural environment. 

  The environment plays an important role in the design of an agent; it, in part, denes how many degrees of freedom an agent possibly can have. In section 2.4, the distinction between a physical, communication, social and task environment is described. An environment allows agents to interact with other agents and at the same time, scarcity of space, time and objects in an environment can create conicts between agents. In section 2.5, several types of coordination mechanisms are discussed that enable conicting agent to negotiate and cooperate. Many situations in our daily life are solved with the help of coordination mechanisms. For instance, trafc lights function because we understand the meaning of the coordination mechanism. Such understanding solves problems that otherwise would occur when several cars want to occupy the same space (crossroad) at the same time. 

  Section 2.6 discusses the applications that are implemented in the eld of MAS. They vary from simple desktop applications forconsumers, towards complicated distributed real-time systems in the industry. Applications of MAS areespecially on the Internetwidespread and have proven themselves a worthy competitor to alternative systems, i.e. due to the Internet, MAS applications are more wanted than ever before and can be the answer to many of today's coordination problems. Finally, section 2.7 closes the chapter with a discussion about what kind of agent and theory is important for our research, i.e. the type of agent will be selected from a variation of agents that differ in the way they are constructed, from simple to very complex. The complexity of such a construction depends on the philosophical ideas behind the design of the agent and its environment, and the amount of complexity we need in order to answer the research questions.（essay代写)

  51Due网站原创范文除特殊说明外一切图文著作权归51Due所有；未经51Due官方授权谢绝任何用途转载或刊发于媒体。如发生侵犯著作权现象，51Due保留一切法律追诉权。
  更多essay代写范文欢迎访问我们主页 www.51due.com 当然有essay代写需求可以和我们24小时在线客服 QQ:800020041 联系交流。-X（essay代写)