Wireless Sensor Networks

Wireless sensor networks (WSNs) have gained tremendous attention during the last years. They consist of small sensing devices capable to communicate with each other within short distances, but in a collaborative manner they can process data on a larger scale. WSNs are usually deployed randomly in a region under scrutiny and have tight energy and bandwith constraints such that the amount of data that can be transmitted and local computations that can be carried out is limited. In most applications, where a large number of nodes is deployed, either placed by hand or perhaps dropped by an airplane, nobody will go to replace the empty batteries which should highlight the need for
energy efficient strategies.

Designed to make inferences about the environment which they are sensing, a major goal is to distribute the computational effort of the used inference algorithms among the nodes to save energy by reducing the communication load for the individual sensors. This is opposed to a centralized approach, where the measurements have to be transmitted to a fusion center (powerful base station) that then carries out the computations. Since many real world spatial-temporal phenomena (e.g. air pressure or temperature fields) tend to be very complex, transmission of the sensor information to a fusion center is impractical in terms of the whole network's lifetime which is due to the limited transmitting range, such that may nodes need to relay information from distant parts of the network to the fusion center. After mentioning not only the interesting properties but also the constraints that WSNs have to cope with, it should be clear why distributed signal processing and in particular distributed learning in WSNs has emerged to an internationally growing research direction.


Sensor Network