Networked sensors - those that coordinate amongst themselves to achieve a sensing task - promise to revolutionize the way we live, work and interact with the physical environment. Fundamental to such coordination is localization, or the ability to establish spatial relationships among such devices. However, existing localization systems such as GPS do not always meet the operational (for example, low power), environmental (for example, indoors) or cost constraints of such systems.
In this talk, I will describe the challenges for localization in very large, ad hoc deployed sensor networks. One approach to localizing small devices is to rely on a system of beacons, nodes that are position-aware by virtue of being pre-positioned, GPS-enabled or endowed with more sophisticated hardware. I will make the case that in unattended sensor networks, instead of relying on extensive pre-configuration, these beacon systems must self-configure, i.e., autonomously adapt to the dynamics of their environmental setting and the availability of other beacons.
I will first quantitatively analyze the impact of beacon density on the quality of localization, and show that localization saturates at a certain beacon density. I will then describe the design of two self-configuring algorithms that build on this observation, to improve localization quality by adding new beacons at low densities or increase system lifetime by rotating functionality amongst redundant beacons at high beacon densities. The talk will include performance results from simulations as well as experimental results from implementation of an RF-proximity based localization system on a variety of wireless testbeds, including tiny devices known as Rene motes, developed at U.C. Berkeley.