Research On A Lightweight Context-Aware Software Architecture

Pervasive computing aims to deeply embed computers in devices rounding human. A harmonious human-computer space is emerging along with the seamless convergence of information space and the real world. The vision is depended on some techniques such as high sensitivity of pervasive software to environment information, nature interaction between human and computer, as well as auto-configuration and adaptation of software and devices. Many technologies such as middleware, mobile computing, interaction, embedded system, sensor and network are used by researcher in pervasive computing area. One of most important research direction of pervasive computing is context-aware computing, i.e. computing system should adapt its behaviors in real time depending on its context and user's needs. In this case, user need not operate computer any more, whereas he supervise computer system which may want a critical instruction from human in some times. In the early 90th of last century, Shilit and Theimer proposed the first context-aware system followed by many interesting research work in this topic. Now it is still an open and attractive area, It poses unique challenges to researchers. The most important and meaningful challenge is to develop adaptable context-aware application.Based on Open Service Gateway Initiative service platform, a lightweight Service-Oriented Architecture, the thesis develops a novel lightweight context-aware architecture LCASOA for creating embedded system in a relatively closed pervasive environment-Star-like Intelligent Space.Firstly, an subject-based context model and a Context Pair language are proposed to support context acquisition, representation, saving and reasoning.Secondly, in this thesis, user delegates goals to guide the automatic assembly of a context-aware application from service components on the fly. Rete-based production reasoning engine resolves a high-level goal by a set of atomic goals during the run-time. Atomic goal is a production rule which can invoke service methods directly.Thirdly, SOA help LCASOA hide the complexity of executive environment to application developer. On the other hide, goal-driven model create a powerful solution for service composition. This thesis also works on the other issues of OSGi such as service dependence, automatic service deployment and service cache management. Fourthly, a goal-driven ubiquitous agent model is proposed. The heart of an agent is a goal planner. Loose-coupled service collaboration and dynamic service binding support mobile agent and agent collaboration and inter-communication.Finally, a prototype for the vehicle information platform is developed to demonstrate the LCASOA's usefulness.