Arquitetura para construção de software adaptativo para redes de sensores

Abstract

This work proposes a systematic approach for adaptation and redundancy employment in wireless sensor networks through a novel architecture. The sensor networks appear as a new computation paradigm. It allows the incorporation of physical data, like temperature, pressure, and light level, to existing software infrastructure. These new oportunities to explore the world also brings interesting challenges. One distinguishing challenge is the autonomic network operation. It means that the network has to work in a totally automatic way, with little or no human intervention. For such situation to become a reality, the application must be adaptive, in the sense that the software must change the execution flow to maintain its services' quality. The change in execution implies in the usage of a new data set, with greater quality, obtained from available redundant resources.The architecture proposes the modeling of an application for sensor networks in abstraction layers, which correspond to successive refinements in the process of understading and organizing this application. In each layer there is a adaptation strategy, which are: role assignment, adoption of alternatives to the roles implementation, specification of execution flows, and inclusion of distributed algorithms. As we specify when and where to use each adaptation strategy during the modeling, we inhibit ad hoc techniques which are usually applied in the project and implementation of sensor networks. For the inspection of services' quality, there is a adaptation data layer shared with all other layers. It contains a set of values that allows the effectiveness' evaluation of both the current adaptation and its alternatives. The architecture viability was analyzed in a case study that involved the modeling of a typical sensor network application.