Sensor Network Localization from Natural Phenomena


Most sensory data has limited utility without location information, and manual node localization becomes impossible for large, inaccessable, or mobile sensor deployments. Thus, autonomous localization is crucial for many sensor-network applications. We are developing a distributed-localization algorithm for the PLUG indoor sensor network by analyzing commonly detected sound, light, and vibration sensory data from naturally occuring phenomena. The system enters active mode when its sensed region stays relatively silent and stable (assumed to be unoccupied). Otherwise, it stays in passive mode, with each node estimating its location by collecting sensory data and comparing it to synchronized data from other neighborhood nodes. In active mode, each node ocassionally generates predefined mimics of natural phenomena such as sonic transients, or manipulates an attached light source. The main features of this approach are distributed properties, lack of heavy infrastructure, unobtrusive exploitation of background phenomena, and application of junction trees for message passing.