Loading...
Please wait, while we are loading the content...
Similar Documents
Fusion de l'information dans les réseaux de capteurs : application à la surveillance de phénomènes physiques
| Content Provider | Semantic Scholar |
|---|---|
| Author | Ghadban, Nisrine |
| Copyright Year | 2015 |
| Abstract | This thesis investigates two major problems that are challenging the wireless sensor networks (WSN): the measurements accuracy in the regions with a low density of sensors and the growing volume of data collected by the sensors. The first contribution of this thesis is to enhance the collected measurements accuracy, and hence to strengthen the monitored space coverage by the WSN, by means of the sensors mobility strategy. To this end, we address the estimation problem in a WSN by kernel-based machine learning methods, in order to model some physical phenomenon, such as a gas diffusion. We propose several optimization schemes to increase the relevance of the model. We take advantage of the sensors mobility to introduce several mobility scenarios. Those scenarios define the training set of the model and the sensor that is selected to perform mobility based on several mobility criteria. The second contribution of this thesis addresses the dimensionality reduction of the set of collected data by the WSN. This dimensionality reduction is based on the principal component analysis techniques. For this purpose, we propose several strategies adapted to the restrictions in WSN. We also study two well-known problems in wireless networks: the non-synchronization problem between nodes of the network, and the noise in measures and communication. We propose appropriate solutions with Gossip-like algorithms and smoothing mechanisms. All the techniques developed in this thesis are validated in a WSN dedicated to the monitoring of a physical species leakage such as the diffusion of a gas |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://www.theses.fr/2015TROY0037/document |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
