Localization and fault detection in wireless sensor networks. (Localisation et détection de fautes dans les réseaux de capteurs sans fil)

In this thesis we have covered three themes related to wireless sensor networks. The first one concerns the detection of measurement errors in sensor readings in a wireless sensor network. In order to identify a faulty node we have used soft computing techniques. A fuzzy inference system and a recurrent fuzzy inference system are used to model a node as far as its sensor measurement is concerned. The sensor measurement of a node is approximated by a function whose arguments are the real measurements of the neighboring sensors. The return of the function is the estimated value of the sensor measurement. The difference between the approximated value from the model and the actual measurement of the sensor is used as an indication for whether or not to declare a node as faulty. Then we focus on the localization aspect of all the nodes in the network. Once the intermediate distances between the connected nodes have been calculated, the task remained to be accomplished is to find the position of all the nodes by using as minimum number of anchors as possible. Thus, we have proposed a localization method that uses exactly three anchor/beacon nodes. The motivation for the proposed localization scheme stemmed from the fact that a plane and hence all points on it are completely described by defining/knowing exactly three points. But how to integrate this idea in relation to localization in wireless sensor network is discussed in the second part, where we are able to attain the estimated position of all sensors by using only three anchors. Finally we have focussed our attention on the power loss in a node signal due to voltage droop in the battery of the node. Since our proposed localization algorithm uses the strength in the signal from different nodes, paying attention to the received signal strength is crucial. When the battery of a node looses voltage, there is a decrease in the signal strength from that node. This decrease of strength can be interpreted as an increase in the distance between the respective nodes. In fact this is a misinterpretation of the RSS from localization point of view. Thus in the first part we propose a method to compensate for the apparent loss in signal power due