Optimizing Cluster Head Election in IoT Networks Using Reinforcement Learning

IoT is a hot cake in all sectors from daily wearable’s to automated industries. They avoid disasters, reduce the manpower, and increase the performance in all fields. But the wireless sensor networks used in IoT are resource constraints and ensuring the security, energy efficiency is a challenging task. This paper addresses the review of the existing routing protocols in IoT. The observed limitations are suboptimal energy usage, resource inefficiency and security threats during communication. To resolve these issues we propose a cluster based WSN, where the clusters emerge and cluster head (CH) is elected based on reinforcement learning algorithm (RL). Initially cluster head is elected using K-Means algorithm and in the second iteration Cluster head is dynamically elected based on important factors like residual energy, distance and delay using RL. The nearby nodes to the CH that doesn’t have obstacle around them were taken into consideration in CH election. A MS is used to achieve energy efficiency. In data transmission process cluster members forwards monitored data to cluster heads, from cluster heads mobile sink will collect the data by reaching all cluster heads. Later it forwards data to end user. There are many applications such as e-health, smart home, face recognition, automated industries etc, are working on the basis of IoT. Our proposed algorithm achieves great energy efficiency through the best obstacle free cluster head selection, when compared to GEEC, EADCR, DNN and TTDFP.