Highly scalable data processing framework for pervasive computing applications

One of the key problems in pervasive computing is enabling the collective processing of sensor data obtained from mobile devices such as smartphones. In this demonstration we present a highly scalable storage and processing framework for pervasive computing applications, enabling various estimation problems to be solved from massive data sets, consisting of measurements from millions of nodes or more. The key to achieving such scalability is the use of linear or sublinear time processing algorithms emerging from statistical and machine learning communities. We focus specifically on spatial and spatio-temporal estimation problems in the demonstration, such as prediction of sensor readings, user densities, or wireless network usage in regions for which direct measurements are not available.