This thesis presents EcoSlend, a power management framework for resource-constrained wireless sensor platforms running a power-hungry application. It exploits some hardware functionalities to optimize the energy consumption of a node. EcoSlend implements several power management mechanisms including time synchronization, threshold detection, and low-power listening to make nodes wake up only when they require. Furthermore, EcoSlend offers several sleep / wakeup policies with different duty cycles. Programmers can select the most appropriate policy according to the characteristic of the application running on the nodes. EcoSlend also provides a battery monitor component called the Power Monitor to help programmers get the battery status during run-time. To ease programmers' burden, the implementation of this framework is packed into several APIs so that programmers can apply EcoSlend to their applications with just through a few function calls. We apply EcoSlend to two “real-running” wireless sensor network (WSN) applications to evaluate the efficiency of our framework. The experimental results show that both applications can save at least 70% of current consumption after applying the appropriate EcoSlend policies with only a few hundred bytes code memory overhead.
本論文提出了EcoSlend,針對資源有限且執行一個耗電的應用程式之無線感測平台所設計的電源管理框架。EcoSlend利用一些硬體的功能來對無線感測器的耗電進行優化,並且實做了幾個電源管理機制,包括時間同步(Time Synchronization)、臨界值偵測(Threshold Detection)以及低耗電封包監聽(Low-power Listening)。 針對不同的工作週期,EcoSlend設計了三種醒睡的策略,使用者可以根據無線感測器上執行的應用程式之特性來套用最合適的策略。另外,EcoSlend也提供了一個電池監控元件來幫助使用者動態的取得目前電池的狀態。為了減輕使用者負擔,本論文的實做都被包成應用程式介面(Application Program Interface)的形式,使用者只需透過數個函式呼叫便可套用我們設計的電源管理框架到無線感測平台上。 為了驗證EcoSlend的效能,我們將此框架套用到兩個實際運作的無線感測網路(WSN)系統中。實驗結果顯示,兩個系統在套用合適的策略之後都節省了至少百分之七十以上的耗電,並且只額外佔用少量的記憶體空間。