Study Of Agent Role Model And Construction Method Of Multi Agent System

Agent theory and technology derived from distributed artificial intelligence, and then was broadly used in other research fields. Agent technology as a new software development method has been gotten more and more attention. As a new abstract with system level, coarse grained, autonomy and flexible, agent provides a brand-new way for analysis and design of complex software, and is called a important breakthrough of software development. Agent technology express a society computing paradigm, so the agent study based on sociology thinking is tendency of agent technology. Agent model and construction based on sociology thinking has formal method and informal method:(1) The formal method studies social mental status and their relationship, use these mental status and mathematical logic to construct social agent model, it has some main problems, such as logic proving difficult, symbolic operation complexity, achievement difficult, etc.(2) The informal method construct agent model based on social organization, social role, etc, and use word, picture, and chart to describe agent model; it has some main problems, such as what is human society paradigm, what is multi agent system model based on social thinking and suitable construction principle of agent, how to construct agent with autonomy and flexible properties, etc. Based on understanding about sociology and social role forming and action theory, this paper uses informal method as study object, studies multi agent system model based on role, agent role model, agent autonomy architecture and agent model technology based on bottom-up way.The main works of this paper are described as follows:1. Study of social role and multi agent model based on role. According to social role forming and action theory, this paper thinks the social role is key line in sociology, summary five features of social role:non-congenital, non-isolated, pattern, duality relationship and single action, establishes human society paradigm based on social role. Through establishment mapping between human society and multi agent system, this paper defines multi agent system model based on role, it includes initial environment, compute environment, role, agent, role interaction, role organization; gives construction principle of multi agent system model based on role:(1) the agent construction has two step, one step constructs autonomy in initial environment, the other step constructs flexible in compute environment;(2) agent interaction model is constructed by role relationship in compute environment;(3) agent organization is constructed by role using bottom-up way according to organization goal. Discussion influencing factors about requirement construction changes of multi agent system, such as resources, agents, software technology, etc.2. Study of agent role and agent architecture based on role. This paper discusses present and main problems about agent role study, bring forward that an agent role is a relationship between agents locate in same environment, and use relational data theory study agent role; use relational algebra to define role formal model, it includes role data structure, role operation and role integrity, and properties of role model indicate that role satisfy module, high cohesion, low coupling; define relationship among role as role dependent, use Armstrong axiom to study role normalization based on role dependent, discuss transitive dependent and derived dependent, and prove role dependent set is a complete set; define role interaction model, discuss autonomy, consistency and persistent properties of role interaction based on transaction theory; define role organization model, analysis dynamic and flexible of role organization; use Pi calculus define a agent architecture based on role, analysis process of role action.3. Study of agent autonomy BDI architecture based on feedback. This paper establishes agent autonomy BDI architecture through simulation autonomy system of human being. The autonomy status of individual is factors in inside environment, summary physiological functions as six systems:external resource operation system, protection system, control system, sense system, effect system and others; establishes mapping between individual autonomy system and BDI model, uses Pi calculus define agent autonomy BDI architecture, analysis about running indicate architecture can achieve autonomy based on feedback, can control inside status and action by itself. 4. Study of agent analysis and design method using bottom-up way. This paper gives agent model technology (AMT). AMT is based on process and model expression of object model technology. AMT has analysis and design two stages:(1) The analysis stage establishes domain model, agent model, role model, interaction model and function model. The domain model is established by domain engineering method, describes environment where agents locate, gain common feature and diversification feature, the common feature define autonomy properties and operations. The agent model describes static structure of MAS, and use class diagram express agent model. The role model describes relationship among agents, and achieve flexible. The interaction model expresses dynamic feature of agent, define script of role action, then gain status diagram of agent, all status diagram compose interaction model. The function model is defined by data flow diagram, and it expresses output value of agent. The operation of agent is gained from interaction and function model.(2) The design stage includes domain design, agent design and organization design. The domain design defines autonomy operation algorithm. The agent design encapsulates autonomy operations, autonomy properties and role. The organization design defines roles, agents in organization, determine store data structure, and deal with concurrent inside or outside organization, etc.The main innovations of this paper are described as follows:1. Based on social role theory, this paper establishes a multi agent system model based on role. This model highlights the individual effect in society environment. Compared with other multi agent system models directly defined by social role concept, this model in this paper can better describe the feature about agent as social role, and better indicate social and organization of multi agent system.2. Based on relational data theory, this paper establishes an agent role formal model. This formal model highlights the effect of role relationship and role dependent in construction agent. Compared with other agent models directly defined by social role concept, this agent model has more clearly boundary and semantic, in favor of agent construction and reuse..3. Based on Pi calculus, this paper establishes an agent autonomy BDI architecture based on feedback. This model highlights feedback effect in keeping balance among autonomy belief, autonomy desire and autonomy intention. Compared with general BDI architecture, this model in this paper limit scope of possible world semantic, architecture is concise and can be easily constructed.Agent theory and technology is a new research filed, either theory or application has more works to do. This paper studies foundation contents of agent model and construction method based on social role, the further work includes:1. Study autonomy architecture of multi agent system. Based on autonomy agent BDI architecture in this paper and social organization theory, study what is autonomy status, autonomy action of multi agent system? how to achieve autonomy of multi agent system? etc.2. Study multi role dependent. Based on agent role dependent in this paper, study what is multi role dependent? How to formal express multi role dependent? What is normalization of multi role dependent? Etc.3. Study measure of agent model technology. This paper only gives informal description of agent model technology, don't use software measure method discussion quality of multi agent system constructed by this technology.