Ambient beacon localization: using sensed characteristics of the physical world to localize mobile sensors

There is a growing need to support localization in low-power mobile sensor networks, both indoors and outdoors, when mobile sensor nodes (e.g., mote class) are incapable of independently estimating their location (e.g., when GPS is inappropriate or too costly), or are unable to leverage localization schemes designed for static sensor networks. To address this challenge, we propose ambient beacon localization (ABL), an unconventional approach that allows mobile sensors to localize by exploiting their ambient physical environment. Ambient beacon localization combines machine learning and free range beacon-based techniques to bind distinct characteristics of the physical world that appear in sensor data of known locations, which we call ambient beacon points (ABPs). Supervised learning algorithms are used to allow mobile sensors to recognize ABPs, i.e., those physical locations that are sufficiently distinguishable in terms of sensed data from the rest of the sensor field. Ambient beacon localization leverages the very same sensed data that nodes are already collecting on behalf of applications. When a mobile sensor finds itself at an ambient beacon point it starts to beacon that location so that other nodes in range of an ambient beacon can localize themselves, for example, by applying existing beacon based localization schemes. In this paper, we present the design of ambient beacon localization and its initial evaluation in a building-sized testbed. Our work is at an early stage but our experimental testbed and simulation results demonstrate that this unusual approach to localization shows promise.