Exploiting semantics for event detection systems

This dissertation is building towards a flexible and scalable middleware for sentient spaces wherein sensors are used to observe the state of a physical environment in real time to create awareness which, in turn, is used to build applications that bring new functionalities and/or new efficiencies to the environment. Examples of such sentient spaces could be as varied as smart buildings that use video sensors for surveillance or for building pervasive applications such as person locator to dynamically instrumented crisis sites wherein sensors carried by first responders and/or brought to crisis sites are used to monitor the crisis site and the progress of response activities. The key goal of this work is towards developing a robust and scalable middleware technology that can handle challenges related to heterogeneity of sensors, limited resources (e.g., network bandwidth), and the need for rapid deployment (e.g., self-calibration of sensing), and programmability (e.g., hiding complexity of sensor and sensor programming from the application writers).