신경망을 이용한 유비쿼터스 컴퓨팅 환경의 통합관제를 위한 시스템 미들웨어 구현

Abstract

Middleware of ubiquitous computing environment is defined as the support of an application in heterogeneous environment, ubiquitous environment. Middleware-the software layer that lies between the operating system and the application-can be defined as software that helps to give users service regardless of lower hardware, operating system, network. In other words, it means software that supports communication between different protocols, system operating systems, databases and applications, and plays a role in allowing applications to operate in any information system environments.

The purpose of this research is to design and implement the system middleware for real-time integration control in ubiquitous computing environment. To accomplish this purpose, firstly requirements for integration control in ubiquitous computing environment was analyzed, and then 3 layer structure and functions of middleware core that compose middleware were derived based on the requirements analysis. The 3 layer structure consists of Application Connector Layer which takes charge of communication connection between heterogeneous information gathering terminal equipment and middleware, External Layer which both guarantees external interface through connector and takes charge of link to internal layer through listener, and Internal Layer which both plays an important role for interface between external and middleware core and takes charge of actual business logic implementation laying on middleware core.

Middleware core manages life cycle of all middleware agents from creation to extinction such as resources management, event management, message management, process management and agent management, etc. It also not only maintains but also manages the whole functions including communications among all modules of systems and resources allocation.

Secondly, after each module of middleware was derived, the whole system structure was designed so that shows how middleware operates with derived modules. At this time, neural network model to sort priority of events by situation was designed. Moreover, the functions of each middleware module was defined concretely. Especially the classification methodology of neural network to handle priority of events in real time in the process of situation recognition of middleware for integrated control is proposed. This methodology enabled to shorten time spent on events priority classification and classify events priority more efficiently without composing rule-database.

This research also implemented the actual middleware system based on the design and offered simple API to enable developer to implement easily even if internal action of system is unknown during developing application program in the environment where this middleware is applied.

Finally, integrated control platform in ubiquitous computing environment was constructed to test performance of the implemented middleware, and actual performance test was executed by composing virtual performance workload examination module and using system workload test tool. The result showed that response time and server resources processing ability are excellent.