Energy Efficient Selection of Computing Elements in Wireless Sensor Networks

A wide range of wireless sensor network applications are characterized by local processing of the sensed data and only meager data communication requirements. Indeed, because sensor nodes are battery powered and wireless communication bears a high energy cost, data transmission can be traded for on-the-node computation to extend node and network lifetime. Furthermore, the energy consumption can be reduced significantly by selecting and realizing the application on an appropriate processing element. In this article, we propose a new statistical technique for energy consumption estimation for a specific application on various platforms. We have empirically verified the methodology on various classes of embedded processors commonly used in sensor nodes. The methodology can also be applied to multiprocessor platforms. Our solution is not only capable to achieve high accuracy but also facilitates the application developer to evaluate different platforms without actually implementing the application on each of these platforms. Our experimental evaluation results for various platforms will help to understand the implications of using different processing elements and their effects on the lifetime of the network.