DISTRIBUTED STRATEGY BASED ADAPTIVE ESTIMATION INWIRELESS SENSOR NETWORK

Abstract

Advancements in wireless sensor networks (WSNs) have revolutionized environmental monitoring, healthcare, and smart city infrastructure by optimizing their functionality and adaptability through adaptive algorithms. This study focuses on implementation and performance evaluation of different distributed strategies for least mean squares (LMS) algorithm based adaptive estimation in WSNs. The distributed strategies considered in this study include incremental, consensus and diffusion strategies. Under the category of diffusion strategies, both combine-then-adapt and adapt-then-combine strategies are considered. Two state-of-the-art variants of the LMS algorithm i.e. beta- Divergence LMS (BLMS), and Kullback-Leibler LMS (KL-LMS), have been included in this study. Hardware implementation of the distributed strategies and variants of LMS algorithm is carried out using readily available sensors and components. Each node is equipped with a micro-controller interfaced with temperature and flame sensors. These nodes employ distributed strategies in real-time to estimate the temperature of a region. Initially simulation study is carried out to assess the performance of the distributed strategies and variants of LMS algorithm. The performance of different algorithms are compared in term of mean square deviation (MSD). In the first case-study, the parameters of the algorithms are set such that the transient performance of the algorithms are approximately similar. This case-study reveals that the BLMS algorithm provides the lowest MSD if there is a constraint on transient performance. In the second case-study, the step size of different variants of LMS algorithm is kept same. It is found that the BLMS algorithm provides the lowest MSD in second case study too. After evaluating the performance of different algorithms through simulation study, these algorithms are implemented in hardware, and the real-time performance is compared in terms of mean square error (MSE).