Research On Intelligent Agent-based Distributed Component Repository

The using of component-related technologies is an effective way to implement software reuse that could increase the productivity of software development, decrease the cost of development and maintainence, and improve the software quality remarkably. So, the research of component-related technologies has become one of the current hotspots of software engineering. Component repository is an infrastructure in software reuse supporting the specification, storage, management, and retrieval of large numbers of components. As the growing of the component repository technology, many research institutions and software enterprises of China have established their practical component repositories.The architecture of traditional component repository is certralized, gathering the components from various sources to a host and managing them by a management system. As the increasing of the number of repositories established by various organizations and the emphasis of protection to the proterty right of components, the requirements of integrating the isolated repositories grow more obvious. The requirements drive us to seek new theories and technologies to implement the integration of isolated and heterogeneous repositoires through some collaboration mechanism and establish a kind of distributed component repository. It will provide a uniform logic view and query interface, help users find the target components from various repositories through the public network, and achieve better sharing of component information. It will finally improve the reuse of software components.This research begins from Intelligent Agent that has been studied in the fields of sociology, artificial intelligence, and software engineering, adopts Agent to implement the distributed heterogeneous component repository, and makes deep research in the system architecture, intelligent mechanism of Agent, and component retrieving methods, etc. The main work and archievements of this dissertation are listed as following:The architecture of the agent-based distributed component repository is proposed for the first time. This dissertation analysed the basic requirements of distributed heterogeneous component repository, made the conceptual analysis and design of the architecture with the method of Abstract State Machine, and proposed the high level logical structure of the ABDC (Agent Based Distributed Component-repository) architecture. Then, the detailed design was educed according to the architecture. The artichitecture provides a logical view that integrates the distributed and heterogeneous repositories, avoiding the high cost of establishing a centralized repository and the heavy workload on the central server. Each member of the system keeps maintaining their own components, which provides a good security. Due to the automony and the intelligence of agents, the agent-based integration solution is flexbile and low-coupling, and is easy to extend.The architecture model of Agents in this system is proposed, and the communicating and knowledge expressing mechanisms of agents are designed. The distributed component retrieval is implemented through the abilities of the agents and the collaboration among them. This dissertation also proposed the concept of Preference according to the fact that some repositories mainly own components of specific domains, established the method of expressing preferences, and designed the actively notifying mechanism according to the retrieval requests based of logical programming technology. This mechanism can help agents understand the preferences of various repositories and actively notify the appropriate repositories when new retrieval requests are submitted, therefore shorten the periods of waiting for repositories the fetch the requests.A self-adaptive approach for retrieval requests optimization based on association-mining is proposed. This research introduced the association mining theory and method in data mining field to component retrieval. With this approach, the agents can obtain the users' implicit retrieval requirements through intelligently mining the retrieval history and user feedback, thus making the retrieval conditions more complete and precise, which can increase the precision of component retrieval. A convertion model of heterogeneous facet schemes is proposed. Generally, different component repositories have specific facet schemes different with each other, and the scheme used in some user's retrieval conditions may also be different. This researchreferenced the matching theory of tree structure, mappinig the facet shceme into tree structure, and modeled the convertion between different facet schemes as the edit operations of tree structures. With the assistant of the convertion, the users can write their requests according to the facet schemes of the local repositories.The research results of the above mentioned theories and technologies have been applied to a prototype system, which validates their feasibility.