Learning-based localization in wireless and sensor networks

Accurately tracking mobile devices in wireless and sensor networks using received-signal-strength (RSS) values is a useful task in robotics and activity recognition. It is also a difficult task since radio signals usually attenuate in a highly nonlinear and uncertain way in a complex environment where client devices may be moving. Many existing RSS localization systems suffer from the following problems: first, many of them are inaccurate. Second, to increase their accuracy, many require costly manual calibration. Third, many of them cannot cope with changing data as users move in a dynamic environment. In this thesis, we will describe our learning-based solution using kernels, manifolds and graph Laplacian for solving the these problems. We will demonstrate the effectiveness of our algorithms for tracking static and mobile devices in complex indoor environments using wireless local area network, wireless sensor networks and radio frequency identification networks with much less calibration effort.