Clock-offset Cancellation Methods for Positioning in Asynchronous Sensor Networks
Paper in proceeding, 2005

For most applications of wireless sensor networks, knowledge about the position of sensors relative to other sensors in the network, or to some global coordinate system, is a key ingredient to successful operation of the network. Estimation of relative node coordinates, based on measured time-of-flight between neighbouring nodes, has been suggested as a means to provide position-awareness in sensor networks where satellite based systems are not available. However, to directly measure the inter-node distances, based on RF or ultra-sound propagation delay, requires the nodes to be tightly synchronized in time. This is an assumption that is not easily justified in sensor networks operating under complexity, latency, power consumption or bandwidth constraints. Joint ML estimation of clock-offsets and node coordinates has been suggested, but, although this approach shows great promise in terms of coordinate estimation accuracy, it does not scale well as sensor networks grow in size. In this paper, we present two linear preprocessing operations that cancels the effect of unknown clock-offsets from the estimation problem. For both operations, we show that the Fisher information on node coordinates in the original data-set remains unchanged after preprocessing. Novel ML estimators of relative node coordinates are proposed, that are of significantly lower complexity, while their performance equals that of the joint ML estimator. © 2005 IEEE.

safe-traffic

Author

Mats Rydström

Chalmers, Signals and Systems, Communication, Antennas and Optical Networks

Erik Ström

Chalmers, Signals and Systems, Communication, Antennas and Optical Networks

Arne Svensson

Chalmers, Signals and Systems, Communication, Antennas and Optical Networks

Proceedings IEEE WirelessCom, Hawaii, USA

981-986
0780393058 (ISBN)

Subject Categories

Computer and Information Science

DOI

10.1109/WIRLES.2005.1549546

ISBN

0780393058

More information

Created

10/7/2017