Abstract
To make applications aware of the environmental contexts is an important topic in ubiquitous computing. Applications with acquired contexts can adjust their status or behavior to minimize user attention and serve people more intelligently. Sensor networks aid applications to extract contexts by connecting various sensors together for context inquiry, propagation, and aggregation. Although sensor networks provide an effective way for supplying context information to applications, their design has to cope with various restrictions, such as energy consumption, communication range, sensor failure, etc. In this thesis, we study the system software design over the sensor network to aid the development of context-aware applications. Without relying on flooding or multicasting to disseminate messages over our sensor network, a novel data-centric routing mechanism, Trailblazer, is adopted. In Trailblazer, every sensor has a set of attribute-value pairs to describe its functionalities. The description of a sensor is hashed to a hash value to identify this sensor. All the sensors maintain certain routing information to help forward messages in the sensor network. Based on Trailblazer, we have developed a framework to aid the construction of context-aware applications on top of sensor networks. The applications can publish or subscribe desired context in sensor networks easily and efficiently. The proposed framework also supports the mobility of sensors. Finally, we demonstrate an object tracking system constructed from the framework.