UCLA

With the explosion of sensing platforms and modalities in recent decades and the growing interest in gathering and disseminating this data came corresponding difficulties in energy management and, more broadly, information management. To combat this deluge of information, there has been considerable research in fields spanning the entire spectrum of computing, from data centers to integrated circuits. In the realm of embedded systems and low power sensing, many have focused on exploiting sensor streams characterized by a very low information rate, i.e. when the data source is highly compressible and thus the phenomenon to be sensed produces a Thin Data stream. This work focuses on several architectures and techniques for reducing power consumption for sensing Thin Data streams. Specifically, the focus of the this work is divided three-fold: (1) We consider an operating systems perspective for allowing duty-cycled scheduling of periodic low power sensing tasks in the face of hardware and temperature variability; (2) we demonstrate low power sensing strategies for sporadic sources in embedded hardware and software with applications in water flow monitoring; (3) finally, we consider low power, low information sensor architectures from a hardware point of view, constructing a low power front-end for compressed sensing.