Research On Model And Methodology Of Adaptive Middleware For Ubiquitous Computing

In 21 st century, with the rapid development of information technology, a significant revolution takes place in the computation model. It is an inevitable trend of computation development from distributed computing to ubiquitous computing. There are three essential elements in ubiquitous computing: embedded, small, and intelligent devices, multiple kinds of persons, and multiple types of smart spaces. Smart space is a fusion of physical world and information space, which provides the users with intelligent services and facilities.Due to the inherent complexity of ubiquitous computing, it poses many new challenges for software middleware infrastructure. It needs urgently the adaptive middleware infrastructure. According to the changes of the run-time contexts, the middleware infrastructure can self-adjust and re-configrate its structure and behavior in order to provide adaptive services for above ubiquitous computing applications. There are two important problems in the research on adaptive middleware for ubiquitous computing: 1) the adaptive middleware model for ubiquitous computing, and 2) the key implementation method of adaptive middleware based on the component technology. It is significant to make research on above two problems. We should solve two problems from both the theoretic and practical views. The thesis aims at the ubiquitous computing and smart spaces. We research on the adaptive middleware model and the key implementation methods, breaking through the difficulties of adaptive middleware implementation during design-time and run-time. The thesis provides an adaptive middleware model and two key implementation methods for ubiquitous computing. The main contents and contributions of our work are as follows.First, we build an adaptive middleware model — SCUD. SCUD model consists of adaptive entities that conform to CCM specification. The elements of the adaptive entity are adaptive component and the composition of these components. The component compositons are adaptive agents, which are dynamic, virtual, and semantic-integrated. SCUD model integrates introspection mechanism based on AOP and reflection, and external-spection mechanism based on the context-awareness technology. To ensure the adaptation in middleware is correct and safe, SCUD model provides a middleware adaptation specification — CMAS that is based on the extended temporal logic.Second, we present SCUD adaptive component allocation method—RIT, achieving the static configurable middleware adaptation during the component assembly and deploy. RIT component allocation method aims at the limited computation resource, and adopts theheuristic strategies. Under the guidance of the CMAS specification, RIT component allocation method solves the problems of computation resource constrains, component interdependences, and component tolerant issues. In addition, RIT component allocation method provides the abilities of dynamic planning and forward checking. The experiment shows that the RIT component allocation method has good performance and provides middleware static configurable adaptation.Third, we propose SCUD component composition method — SEDCC, achieving the dynamic mutable middleware adaptation during the component run-time phase. The SEDCC component composition method aims at correct, safe, and effective component composition during middleware run-time. Under the guidance of the CMAS specification, SEDCC component composition method 1) does not break down the former component interdependences;2) does not interrupt the events communications and interface use among the components;3) ensure the cost of the component composition is minimum. The experiment shows that the SEDCC component composition method has good performance and provides middleware dynamic mutable adaptation.Last, we implement an adaptive middleware — SCUD Ware based on the SCUD model, SCUD component allocation method, and SCUD component composition method. SCUDWare achieves the static reconfigurability, and dynamic mutable middleware adaptation. We use the smart vehicle space as a scenario to make research on ubiquitous computing. SCUDWare middleware is applied into the smart vehicle space, and we develop a mobile music system in smart vehicle space. The experiment shows that the SCUDWare middleware provides static and dynamic middleware adaptation, and can support the requirements of ubiquitous computing. In addition, SCUDWare has good performance for open, dynamic, and various computation environments in ubiquitous computing.This thesis was supported in part by the 863 National High Technology Program, and in part by the Key Program of Natural Science Foundation of China, and in part by the Program for New Century Excellent Talents in University.