Fusão de dados em redes de sensores sem fio.

Abstract

This work provides a general discussion for information fusion in wireless sensor networks (WSNs), allowing us to identify open issues and understand the requirements and the implications regarding information fusion and the resource-constrained WSNs. In this discussion, we survey the state-of-the-art about information fusion inWSNs. By assessing the architectures, models, and methods of information fusion identified in the survey, we propose a framework, called Diffuse, that comprises the main functions and activities of a general fusion process and a specific API that implements useful algorithms for WSNs. The Diffuse framework is a helpful tool thatallows the designer to reason about what types of information fusion, what methods should be used, and how they should be used to accomplish an information-fusion task or application. Although the applicability of Diffuse is ample, as a proof of concept, we show how it can be used to achieve energy-efficient reliability in tree-based routing protocols. Results show that our approach efficiently avoids unnecessaryrouting topology constructions. In some cases, the traffic overhead generated by this approach is 85% smaller than the traffic generated by classical algorithms. In addition, we introduce a routing strategy, based on a role assignment algorithm, to support an information-fusion application. In this case, we consider that WSNs applyinformation fusion techniques to detect events in the sensor field, and propose a role assignment algorithm, called InFRA, to organize the network only when events are detected. In a nutshell, InFRA is an event-based role assignment algorithm that tries to reactively find the shortest routes (connecting source nodes to the sink) that maximize data aggregation. Results show that, in some cases, the InFRA algorithmuses only 70% of the energy spent by other tree-based routing algorithms that are commonly used in WSNs