| In the past few years the field of pervasive computing has expanded and infiltrated down to everyday consumer activities. Smartness of devices is constantly increasing. Smart products are widespread, not just limited to appliances and electronics. With the availability of such products, there becomes an increased need for scalable smart environments that are capable of integrating them to harness their combined benefits. Sensing a space to find out about its devices and opportunities they offer is a particularly important capability in smart spaces. For example, it is very useful to be able to discover the location of a certain device, reason about its role and possible participation in a certain application, and finally be able to control it remotely. Such self sensing and delegation of control is needed by many applications like remotely pinpointing the location of a certain device that is causing some kind of threat. Self sensing requires world modeling when physical world entities are mapped into model artifacts. In this dissertation, we introduce three approaches to creating useful world models. First, we present SensoBot, a mobile sensor platform that has the ability of mapping the space, creating the respective floor plan, locating important landmarks, and approximately locating furniture. In addition, we discuss the implementation of smart plugs, a novel way to locate and control smart devices within a real smart house.;We then propose a novel approach in which classical computer vision algorithms are empowered by opportunities presented by the pervasive space. Our approach, which we call PerVision, extends classical object recognition and tracking algorithms by adding a self-assessment/adjustment loop in which sensors and actuators of the pervasive space are used to vary scene parameters to minimize errors in the recognition process. We present the PerVision concept and algorithms in the context of locating and tracking dumb objects such as furniture in a smart house. Collectively, SensoBot, Smart Plugs, and PerVision take us a few steps closer to realizing the ambitious vision of completely self sensing spaces. |
