Research On Key Enabling Technologies For Smart Space

Pervasive Computing (also known as Ubiquitous Computing) is a new wave of computing paradigm. Its main ideas include 1) the fusion of cyberspace and physical space, 2) the distraction-free computing. Smart Space embodies Pervasive Computing in the scale of a room or a building. It is a working or living environment with embedded computers, information appliances, and multi-modal sensors, allowing people to perform tasks efficiently by offering unprecedented information access and computing assistance. This thesis explores the two enabling technologies for Smart Spaces - the software infrastructure and the context service.The software system of Smart Spaces distinguishs ifself from traditional distributed systems in that: 1) they should deliver functionality in our everyday world, which is a complex and dynamic changing environment; 2) the system will consist of many wireless networked devices, while wireless connection is prone to transient failure; 3) the interaction points for the people are distributed and multimodal, which imposes tight temporal constraints on the messaging between related modules; 4) modules coming from diversified domains need to cooperate, which raise the need for an abstraction model at a proper level for the construction of the system.Currently general-purpose distributed computing software infrastructure or known Smart Space oriented solutions do not sufficiently address all the above-mentioned issues. This thesis presents a new solution called the Smart Platform, which better tackles the problems by: 1) a runtime environment discovery mechanism which allows a new device to automatically find and join the runtime environment; 2) the loose coupled Publish-Subscribe inter-module coordination model, which caters for the spontaneous interaction between modules; 3) a global-local two-layer runtime environment structure. The transient failure of the wireless network can be recovered at the local environment layer. The asynchronous nature of the interactions in the Publish-Subscribe model also prevents modules from blocking on the transient network failure; 4) distinguishing two classes of communications with different QoS requirements, the message-oriented and the stream-oriented. Each is served with a specifically optimized communication layer implementation; 5)a multi-agent system abstraction model, as a powerful tool for integrating modules from different domains. Context aware computing is the effort for computer systems to sense and leverage the rich context information in the physical space. It has been recognized as a promising approach to reach the goal of invisible computing. A context service in a Smart Space provides a centralized point of access to context information for applications in the space. Previous works demonstrate two significant disadvantages: 1) the context access interface is procedural, which means the application should be aware of the arrangement of the context data storage. 2) the context reasoning mechanism is more or less ad-hoc, especially in the case of reasoning with the uncertainty in the context. The key to these issues is a proper context representation and reasoning model, which should satisfy the following requirements: 1) it should have the capability to deal with uncertainties, which is an intrinsic property of context; 2) it should be first-order, since we need to discuss at the level of entities, attributes and relations. This property is also necessary for representing generally applicable context reasoning knowledge. 3) it should have reasonable complexity, since the application wants to now the up-to-date context. Among the few proposed context representation and reasoning models, none can support all these requirements.We propose a novel First-order Probability Logic based solution, Context Logic. It deals with the uncertainties in the context based on the theory of probability, while at the same time, poessing the first-order property. The main difficulty of First-order Probability Logic is the relatively high computation compl...