Optimized Energy-Efficient Machine Learning-Based Clustering and Routing in Wireless Sensor Networks

Abstract

WSNs continue to struggle with the issue of energy usage, network lifetime, and dependable data transfer, especially in heterogeneous networks where nodes have diverse computational and energy resources. The classical clustering algorithms are usually characterized by uneven distribution of cluster-heads, inefficient routing paths, and poor responsiveness to changes in network dynamics. This article presents a new framework, OEEMLCR (Optimized Energy-Efficient Machine Learning-Based Clustering and Routing), which is a combination of Particle Swarm Optimization-based K-means clustering and Coati Optimization Algorithm-based routing and Q-learning adaptation. The suggested methodology fills in crucial gaps through the use of a dual-objective fitness function that both maximizes the compactness of space and the homogeneity of energy when forming clusters and introduces the learning technique of reinforcement to allow adaptive routing behavior under the influence of experience gained in the network. Experimental validation of heterogeneous network situations show that OEEMLCR can attain significant gains over existing protocols: network lifetime is increased 13.6% over baseline machine learning methods, cumulative packet delivery is increased 108%, energy efficiency is increased 340%. The framework operates in a stable way during 1000 rounds of the simulation and is 96.7 percent of the initial network power, far outperforming LEACH, DMHT, EDMHT, and EEMLCR in lifetime, throughput, percentage ratio of delivering packets, and the network resilience criterion.