Methodology Overview of Agent
1. Introduction Regardless of whether or not the intelligent main body (AO: Agent-Oriented) method is like a data stream (DFO: DataFlow-Oriented), the data structure (DSO: DataStruct-Oriented), object-oriented (OO) : Object-oriented) Behind the new generation of software development methodology. With the continuous improvement of software system service capacity requirements, the introduction of intelligent factors in the system is inevitable. Theagent is an important and advanced branch of artificial intelligence research, causing science. , Engineering, technical community. Stanford University Barbara Hayes-Roth mentioned in the special report of IJcai'95: Intelligent computer main body is both artificial intelligence initial goals. The network leads to network computing The appearance of this concept. Today, parallel and distribution calculations are highlights of network computing. As a factor of parallel, distributed calculation, the partner, the target naturally produces MAS (MULTI-Agent System), MA (Mobile Agent). Direction. All scientific and technical achievements will greatly improve the ability of the software system, follow the needs of the times. Previous methods cannot support new development models, must match the development method. Agent Methodology of the times.
2.Agent and agent-based System Overview Agent's research dates back to the 70-year distributed artificial intelligence (DAI: Distribute Artifical Intelligence), mainly divided into two research routes: a manuscript of the classic manual intelligence, the main research agent Behavior, multi-agent negotiation model, etc., its research direction can be divided into proxy theory, agent architecture, proxy language, multi-agent system, etc., some computer scientists are called "intelligent agents" or strong definition agents; another Around the 1990s, we can weaken the application based on the application, and the strong definition of the classic artificial intelligence will weaken, broaden the application range of the agent, the new research direction mainly includes the agent interface, the agent based software engineering (AOSE). AGENT features : 1. Action on behalf Others: Acting has the ability to represent others, that is, they all represent users. This is the first feature of the agent. 2. AutoMY: A agent is an independent computing entity It has different degrees of self-control ability. It can solve practical problems in non-prior planning and dynamic environments, independent discovery and request resources, services, etc. in accordance with non-user participation. 3. Active : The agent can follow the commitment to take active, representing the target-oriented behavior. For example, the agent on the Internet can roam all the net, collect information for the user, and submit information to the user. 4. Reactivity: Agent: agent can perceive The environment is appropriately reacted. 5. Social Ability: The agent has certain sociality, which may communicate with users, resources, other agents representatives representatives. 6. Intelligence (Intelligence): The agent has a certain degree of intelligence, including a series of smart behaviors such as self-learning. Agents may express other properties to some extent: 7. Callaboration: More advanced agents can work together with other agents to work together Single agent cannot be completed. 8. Mobility: Moved (Mobility): Move ability, to complete the task, can move from one node to another. For example, access the remote resource, transfer to the environmentally suitable node for work, etc. Honesty, compliance, IL, etc. Because of Agent's characteristics, Agent-based systems should be a superior system that integrates flexibility, intelligence, scalability, robustness, organism, and more.
3. OA Method Next We focus on system decomposition, organization to analyze OA methodology support for modern software systems. 1) Systematic analysis facing Agent's system analysis is an effective way to split complex system problem space. Object-oriented system The key to the oriented design is to some extent to map the real world entity. And Agent is more reflected in reality than the object. It not only abstracts the entity's characteristics, the action, and feelings, mind, promise, etc. This is from real. It is easier to break the system into a flexible, strong interactive system in Agent. On the one hand, due to the self-made, reactivity, the subsystem has its own control flow. And in parallel, distribution has become systematic Trend, Agent can undertake multiple control flow in this system, reducing system analysis complexity. The base is unpredictable (this system often has high spirituality, high processing capability) processing capacity system advantage over I: This system characteristic is one of the important features of the modern software system. It reflects the intelligence of the system. In the process of decomposition, the relationship between the subsystem can also be mapped to the interaction between Agent. Collaboration. 2) Contains an agent organization From a mapping perspective, agent-oriented system organization is more suitable for the dependencies and interaction relationships in the system. Complex software systems often contain complex tissue relationships, these organization relationships make some scattered modules It is divided into a conceptual entity; secondly, there is a high-level link relationship between these characterized entities; again, the relationship between the entity is changing, requiring each entity to have this change Ability. And Agent-based software system has a flexible organization of relationships, which can be automatically established and released according to a certain mechanism. This structure allows individual Agents to be independently developed, then add to the system. In this way, based on Agent-based complex software systems There is a good growth. Compliance with the requirements of the software system for system growth. 4. System modeling for Agent 4.1 GAIA Modeling Method GAIA model is proposed by people such as WoolDridge, Jennings, and Kinny. Suitable for an Agent in the closed system Analysis (AOA) and Design (AOD) .Gaia supports the structure of the Agent and the modeling between the social and organizational structure between the Agent, it requires a single Agent's capabilities and interactions between Agents is static during system operation. This makes the Agent lose its own characteristics. The main process has two phases of analysis and design: analysis process focuses on the understanding of the system and system structure, mainly involving roles and roles in Agent Organization. The interaction between the design; mainly defines the actual structure of the Agent system, mainly involving the constituent system AG ENT class and instance, as well as the services provided by each agent and familiarity. Note: For details, please refer to WoolDridge MJ, et al. "The Gaia Methodology for Agent-Oriented Analysis and Design" [J]. Autonmous Agents and Multi- Agent Systems, 2003 3 (3): 285-312
4.2 Multi-Agent System Modeling Method GAIA Modeling is not suitable for systemic systems with a certain social guidelines. Therefore, Wood and del Oach extension GAIA proposed Multi-Agent System Engineering Method (Multiagent Systems Enginneering) Methodlolgy), it supports the creation of dynamic code, the target is to convert the initial system specification into an Agent system that can be operated. The development process describes the following: 1) Get the target of the system. Convert the initial system specification to structured system objectives. 2) Converting the structured target to a more useful role in a multi-Agent system. 3) Application example. Define logic paths between different systems or interior of different roles within the same system. 4) Establish an Agent class. The role is mapped to the specific Agent class map. 5) Establish an Agent session. Define the communication protocol between the Agent. 6) Assemble the Agent. 7) System implementation. Instantiate the Agent class into an agent.
NOTE: See details: A Omicini, et al "Agent-Oriented SoftWare Engineering for Internet Application" [M}, Franco Zambonelli, Published as Chapter13 in the book:. Coordinaton of Internet Agents: Models, Technologies and Applications, 2000.4.3 The mode of modeling of information systems is proposed for Agent's information system modeling WAGNER, and Agent-Object Relationship, AOR) .aor model is based on currently widely used entity relationship model (ER) and relational database The RELATIONAL DATABASE MODEL, RDM .ER model is mainly simplified to convert contact between entities to specific data system design, which supports static entities and object modeling. But does not apply to information systems The dynamic entity or agent.AOR model is used to make up for the deficiencies of the ER model, and provide relationship between Agent. In AOR, there are six types: intelligent (Agent), event, behavior, obligation ( Commitment, statement. The obligations and statements are usually paired. An Agent's obligation is to be seen as other agents. The organization is built into a group of agents. Each Agent has the right Complete specific behaviors, but also assume certain obligations, such as monitoring the statement related to the Agent organization. It is advantageous for the Agent system than the "divide" strategy of traditional software engineering classics. It is, Agent can be flexible. Context-related methods rather than interact with external interactions through some predetermined interface functions. Soon, from scientific research to application system deployment, intelligent systems is an inevitable trend. And Agent-based systems are the core. Come AOP, AOSE is inevitably developed. However, it is clearly specified, implemented, and verifies the Agent-oriented system thus makes Agent-based projects that are absolutely a difficult way to explore.
