MultiAgent-Based Nonlinear Edit Cooperative System

Agent is an important research field developed in recent years. Learning and cooperation of agent are key issues as the foundation of adaptive Multi-agent systems, which are significant to the developments of both theories and applications of Multi-agent systems. The research points of this thesis include the theories, learning and cooperation models and application of Multi-agent systems in Nonlinear Edit.The thesis comprehensively analyzed the description theory and computational theory of Multi-agent systems. The thesis holds a different view towards using Cantor's diagonal argument to compare the cardinals of two infinite sets. The thesis proposes an infinite set to be defined as X_g. X is the set of initial elements. g is the produce mechanism. It's pointed out that Cantor's continuum hypothesis is rather about the complexity than about the cardinals of infinite sets. As for logic, due to its limitation, it is necessary to introduce new principles for further research in Multi-agent systems. The combination of natural scientific principles and classical algorithms becomes the new computational foundation. However, those theories are bottom-up and cannot explain or predict systematic behavior, so emergent behavior is still a difficult but hot issue in the field of complex Multi-agent systems.The thesis comprehensively analyzed learning models and cooperation models of Multi-agent systems, which points out how to learn and cooperative according to actual problems and application environments. Learning is a procedure for agent to acquire knowledge and experiences, improve performance, discover laws or principles and adapt to environment. Cooperation is the key to ensure agents working together.Comparison and analysis of the current models also illustrates the importance of agent-environment interaction. Those learning models can be classified into belief-based learning models and performance-based learning models. The former is mainly based on classical theories while the latter mainly on non-classical theories. And so do cooperation models. A efficient model is pertinent to its application domain.The thesis proposed a new MPEG-7-and MultiAgent-based Nonlinear Edit System which uses a Generic model to describe multimedia because MPEG-7 provides standard, comprehensive and flexible multimedia description tools that enable scalable, intelligent and interoperable multimedia applications. The generic model is illustrated and exemplified by Extensible Markup Language. It aims at clarity and flexibility to support MPEG-7 applications such as edit. It is easier for agent to acquire knowledge from MPEG-7 multimedia documents, thus agent can domore content-based work such as classification and query, even task assignment.In MultiAgent-based Nonlinear Edit System, agent and human learn from each other and cooperate to edit multimedia documents. The learning models include probabilistic learning model, profit-sharing learning model and go-with-the-flow learning model. Cooperation models include contract net model and assignment model. In order to characterize agents in a personal manner, the thesis proposed to endow agents with expressions according to their internal states.It is pointed out that the expansibility of MultiAgent-based Nonlinear Edit System is better than those of some existing software such as Final Cut Express, Avid and Premiere. New functions or agents of new types are able to add into the system. Such as Internet Search Agent which can collects more multimedia information from internet. The main contribution includes:(1) The comprehensive analysis of foundational theories, learning models and cooperation models of Multi-agent systems. Comparison and analysis of the current models illustrates the importance of agent-environment interaction. The proposal of the conception: complexity of set.(2) The proposal of a cooperative model of MultiAgent-based Nonlinear Edit System-CNLES. Through learning and cooperation of agents, the model provides a new way to edit compared with the existing nonlinear edit software such as Final Cut Express, Avid and Adobe Premiere.(3) The proposal of a new MPEG-7 based Generic Model to describe multimedia that enable scalable, intelligent and interoperable multimedia applications such as edit. The proposal of Database Agent which can classify and query MPEG-7 multimedia documents on the basis of their content.(4) The proposal of Task Agent which can cooperate with Database Agent, and divide edit tasks based on the content of multimedia, then assign tasks based on contract net model. Task Agent enables distributed parallel processing of the system;(5) The proposal of Personal Agent which learns from and cooperate with human. Expressions are endowed to PA in order to characterize agent in a personal manner.(6) The expansibility of CNLES. New functions and new agents can be added into such a system to improve the performance of the system.