Publica
Hier finden Sie wissenschaftliche Publikationen aus den FraunhoferInstituten. Harnessing interference for analog function computation in wireless sensor networks
 IEEE transactions on signal processing 61 (2013), No.20, pp.48934906 ISSN: 00963518 ISSN: 00189278 ISSN: 00961620 ISSN: 1053587X 

 English 
 Journal Article 
 Fraunhofer HHI () 
Abstract
It is known that if the objective of a wireless sensor network is not to reconstruct individual sensor readings at a fusion center but rather to compute a linear function of them, then the interference property of the wireless channel can be beneficially harnessed by letting nodes transmit simultaneously. Recently, an analog computation scheme was proposed to show that it is possible to take the advantage of the interference property even if nonlinear functions are to be computed. The scheme involves some preprocessing on the sensor readings and postprocessing on the superimposed signals observed by the fusion center. Correspondingly, this paper provides a thorough base for a theory of analogcomputing functions over wireless channels by specifying what is the maximum achievable. This means it is determined for networks of arbitrary topology which functions are generally analogcomputable over the channel and how many wireless resources are needed. It turns out that the considerations are closely related to the famous 13th Hilbert problem and that analogcomputations can be universally performed in the sense that the preprocessing at sensor nodes is independent of the function to be computed. Universality reduces the complexity of transmitters and the signaling overhead, and it is shown that this property is preserved if nodes leave or join the network. Analogcomputability is therefore of high practical relevance as it allows for an efficient computation of functions in sensor networks.