Energy-Efficient Dynamic Clustering for IoT Applications: A Neural Network Approach
Li Manman, Qin Xin, Pratik Goswami, Amrit Mukherjee, Lixia Yang
Abstract
The Internet of Things (IoT) realizes the interconnection of different paradigms in information technologies along with their connectivity. With its evolution, the cost and energy efficiency along with ease of life stands as a challenge towards its deployments. Advances in Artificial Intelligence coupled with IoT connectivity provide lots of potentials for real-time communication applications. To address the issue of energy-efficient computing for these applications, we propose an improved dynamic clustering algorithm which can be implemented in IoT applications which comprises of heterogeneous Wireless Sensor Networks (WSNs). Initially, we use neural network and Copula theory to process the information quantity based on power demand by individual clusters. This avoids the information redundancy and the waste of resources caused by repeated construction of similar type of clusters (as in conventional methods). According to the power requirement, the nodes based on the applications are divided into two initial clusters, and compared with the set thresholds. These are then used to assign logical values to the nodes in the cluster. Since, the amounts of observation information of the nodes are not always useful due to random energy consumption; the Back Propagation Neural Network (BPNN) is used to optimize the amount of information to form the final dynamic cluster efficiently. The simulation results showed that the proposed method can effectively utilize the information in the cluster and balance the inter-cluster cooperative communication energy efficiently for IoT applications. The effectiveness of the proposed approach and the superiority compared with the traditional methods had also been demonstrated through the simulation results. We hope our work can further stimulate the investigations on energy and cost efficient methodologies for smart IoT applications.