Energy Efficiency and Reliability in Dense Sensor Networks

This research addresses some important components in the theoretical and algorithmic signal processing machinery needed to make low-power, ubiquitous sensor networks a reality. The physical and hardware attributes as well as the computing and communication capabilities of these low-power, low-cost sensors, particularly those based on high-density low-cost MEMS devices, have the potential to revolutionize next-generation information technology. Next-generation MEMS sensors are expected to be very cheap and very small (of the order of one millimeter cube) with a communication range of several hundred meters and a bandwidth of tens to hundreds of kilobits per second. The challenge is to build a pervasive, reliable, massively
distributed, dynamically self-configuring dense sensor network system out of these low-cost, ubiquitous devices.
The challenges presented by these networks are far beyond existing theories and algorithms, and in many cases require a fundamental paradigm shift from centralized to distributed architectures. Reliable centralized high-performance computing platforms need to give way to a bank of distributed miniaturized, inexpensive, easily deployable, and individually unreliable component nodes which, as a group, however, are required to be robust, energy-efficient, and capable of far more complex tasks. This research program will develop some important components of signal processing and communication system machinery to realize these networks. The focus is on the important components of bandwidth- and energy-efficient, reliable, and robust compression and transmission of sensor network data in a fully distributed fashion. It explores both the theoretical foundation of the relevant multi-terminal settings of this paradigm, as well as computationally efficient distributed processing algorithms aimed at narrowing the gap between theory and practice. Strategies will be developed for optimal compression/transmission for sensor networks where the key abstraction is the use of cooperation but not communication among the sensors to maximize energy-efficiency.