Publication: Procesos de decisión de Markov para comunicaciones selectivas con distribuciones de importancia dependientes
Loading...
Identifiers
Publication date
2013-10
Defense date
2013-10-30
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
Las redes de sensores inalámbricos son una tecnología prometedora que
conlleva necesidades específicas al respecto de gestión de recursos energéticos.
En este proyecto se ha afrontado el diseño de estrategias de transmisión selectiva
para redes de sensores inalámbricos utilizando secuencias de mensajes
caracterizados por importancias con dependencia estadística entre importancias
de mensajes consecutivos, modelando las decisiones del nodo mediante
un Proceso de Decisión de Markov. Se ha observado el comportamiento selectivo
óptimo según dicho modelo para distribuciones simples, identificando
patrones repetidos para energía alta y deduciendo la expresión teórica para
dicha tendencia asintótica.
Se ha aplicado dicho modelo a dos situaciones distintas: importancias
consecutivas con dependencia estadística que se transmiten o descartan e importancias
independientes sobre las que la decisión es transmitir o acumular
en el nodo. Finalmente, se ha tratado de proponer soluciones subóptimas
eficientes, implementables con información estadística incompleta y se ha
comparado la e ciencia de estas estrategias frente a la óptima y frente a
nodos no selectivos.
Wireless sensor networks are a promising technology that entails specific needs regarding energetic resources management. In this proyect, the design of selective transmision strategies has been addressed in the context of nodes in a sensor network which receive messages characterized by a measure of its importance or utility to the network, considering there is statistical dependence between consecutive importances, and modelling the node decision by means of a Markov Decision Process. The optimal selective behaviour on this model has been observed for simple distributions of importances, identifying patterns for high energy and deriving a theoretical expresion for that asymptotic tendency. Two di erent cases have been studied with this model: statistical dependence between importances of messages arriving to the node which are either transmitted or discarted and independent importances that are transmitted or accumulated in the node. Finally, we have tried to propose suboptimal efficient solutions, which are easy to implement on real nodes with incomplete statistical information, comparing the behaviour of this suboptimal strategies with the optimal one and not selective nodes.
Wireless sensor networks are a promising technology that entails specific needs regarding energetic resources management. In this proyect, the design of selective transmision strategies has been addressed in the context of nodes in a sensor network which receive messages characterized by a measure of its importance or utility to the network, considering there is statistical dependence between consecutive importances, and modelling the node decision by means of a Markov Decision Process. The optimal selective behaviour on this model has been observed for simple distributions of importances, identifying patterns for high energy and deriving a theoretical expresion for that asymptotic tendency. Two di erent cases have been studied with this model: statistical dependence between importances of messages arriving to the node which are either transmitted or discarted and independent importances that are transmitted or accumulated in the node. Finally, we have tried to propose suboptimal efficient solutions, which are easy to implement on real nodes with incomplete statistical information, comparing the behaviour of this suboptimal strategies with the optimal one and not selective nodes.
Description
Keywords
Redes de comunicación, Redes de sensores inalámbricas, Procesos de Markov, Programación estocástica, Optimización energética