An improved method of Cluster Head Selection using Machine Learning in WSN

Abstract

Wireless sensor networks (WSNs) have recently piqued the interest of researchers because of its significant part in a number of apps such as environmental tracking, smart homes, and health care monitoring. Energy consumption is a key issue in WSN that has a direct impact on the network's performance because nodes' energy sources are restricted. For extending the lifetime of WSN networks, clustering is one of the most effective topology control strategies. However, because data from across the region is routed through them to reach the base station (BS), the cluster heads (CHs) nearest to the BS in the network are overwhelmed. As a outcome, their energy is lost rapidly, as well as system is split into many portions. The right CHs must be chosen to enable scalability, higher coverage, and energy efficiency in large-scale WSNs. As a result, the motive of proposed study is to include an energy-efficient grid clustering-based routing algorithm in the WSN protocol to boost the network's vitality. In the initial stage, the grid approach was employed to split the nodes into effective cells. One or more CHs are chosen in each grid depending on average throughput and residual energy. Chain routing is utilized for inter-cluster communication during the data transmission phase, in which each CH delivers total information to the BS via adjacent CHs. Moreover, the proposed protocol reduces clustering frequency, reducing the amount of energy required in overhead control packets as well as processing time. The results show that the suggested clustering methodology outperforms in means of throughput, PDR, as well as delay analysis. And the lifespan of the system in the WSN will be improved. When compared to existing clustering-based routing protocols, simulation research shows that the suggested protocols improved network throughput, PDR, and communication delay by 3% to 4%.