Energy efficient Computational Data Offloading for Prolonging Lifetime of WSN: A Machine Learning Framework

Abstract In a cluster based hierarchical wireless sensor networks (WSNs), multiple sensor nodes are grouped together to build a cluster. Here, a cluster head (CH) node is selected to perform the computational task for aggregating the sensor collected data and forward the aggregated data towards a network gateway for further analysis. Thus, the computational data offloading from CH node is an important task for designing an energy efficient WSN. In order to improve the energy efficiency of the WSN, a computational data offloading technique is proposed in this paper under hierarchical arrangements of the sensor nodes. The proposed approach distributes the task of computational data aggregation into some other member sensor nodes, instead of only relying on the CH node of a cluster. In this proposed approach, the relevant component analysis (RCA) is applied to learn intra-cluster distance metric, based on the selected features of sensor nodes. After performing RCA, the distance metric learning (DML) approach is used to find a nearest neighbour node from a sensor node to aggregate the collected data, which is either a CH or any other member sensor node placed closer. To connect any other member sensor node with any sensor node, a temporary link has been created to find a route to reach data towards the nearest sensor node. The proposed work improves the energy efficiency of the network, and that in turn can prolong the WSN lifetime. Various simulation results show the superiority of the proposed approach over state-of-the-art works.