Variable Step Size LMS Algorithm for Data Prediction in Wireless Sensor Networks

Wireless communication itself consumes the most amount of energy in a given WSN, so the most logical way to reduce the energy consumption is to reduce the number of radio transmissions. To address this issue, there have been developed data reduction strategies which reduce the amount of sent data by predicting the measured values both at the source and the sink, requiring transmission only if a certain reading differs by a given margin from the predicted values. While these strategies often provide great reduction in power consumption, they need a-priori knowledge of the explored domain in order to correctly model the expected values. Using a widely known mathematical apparatus called the Least Mean Square Algorithm (LMS), it is possible to get great energy savings while eliminating the need of former knowledge or any kind of modeling. In this paper with we use the Least Mean Square Algorithm with variable step size (LMS-VSS) parameter. By applying this algorithm on real-world dataset, we achieved maximum data reduction of over 95% for star topology and around 97 % when data aggregation was taken into account for cluster-based topology, both for error margin of 0.5°C. Using mean square error as metric for evaluation, we show that our algorithm outperforms classical LMS technique. [PUBLICATION ABSTRACT]