Data Reduction Algorithms based on Computational Intelligence for Wireless Sensor Networks Applications

Abstract

Wireless sensor networks (WSN) are critically resource constrained due to small memory, limited power supply, low processing capability and narrow communication bandwidth. Tremendous researches are geared towards optimizing some aspects of packet transmissions to mitigate those constraints. The energy efficiency of a sensor node is affected by the process of data packet transmission from the sensor board to the fusion center (FC) and also by its packet size. An effective technique to reduce data transmission within the WSN, is to locally reduce the number of packets before transmission. In this paper, the performance of different computational intelligence based algorithms that reduce the data packet traffic is presented. These methods are data reduction based on artificial neural networks (DR-ANN); data reduction methods based on Independent Component Analysis (DR-ICA) and one that is based on regression utilizing deep learning method (DR-GDMLR). These algorithms have been applied to different applications and datasets type. The simulation results with best performance is shown by the DR-ANN algorithm that reduced the size of transmitted data by 66%, while the other two algorithms only reduced the size by 33% only.