| Summary: | 碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 95 === Digital life is no longer a future and is being realized via a lots of research results on ubiquitous computing. In ubiquitous environments, users can access ubiquitous services in an “anyone, anywhere, anytime”, manner. According to the service requirements, each ubiquitous service is carried out by the cooperation among a set of devices, called device composition. Thus, it becomes an important and complex problem to arrange resource-limited devices in bandwidth-limited ubiquitous networks for all requested ubiquitous services with optimized user-demanded service quality. This optimization problem is called Resource-constrained Device Composition Problem (RDCP).
In this thesis, we first formulate RDCP with mathematical models. In our work, RDCP has been reduced to Multi-dimension Multi-choice Knapsack Problem (MMKP) by transforming the model of RDCP into the model of MMKP. Thus, MMKP algorithms can be directly applied for solving RDCP. Moreover, we define a service quality utility function to quantify the quality of a device composition which depends on the user-defined QoS factors, such as reliability and availability, of each device in this device composition. Thus, users could set their preferences on service quality by configuring the weights of QoS factors in defined service quality utility function.
We also propose Service-Oriented Device Arrangement System (SODAS) to automate the service-oriented device arrangement process. In SODAS, Ubiquitous Environment Database (UEDB) is designed for preserving the statuses of ubiquitous services and ubiquitous networks; UEDB Manager (UEDM) is the interface to access UEDB; and Service-Oriented Device Composer (SODC) is the core subsystem to solve RDCP through MMKP algorithms.
Finally, the results of computational complex analysis and performance evaluation show that SODAS is capable of arranging resource-limited devices in bandwidth-limited networks for all requested ubiquitous services. Moreover, the various user-demanded service quality preferences are achieved according to the configured weights of QoS factors in defined service quality utility function.
|
|---|
