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

The use of Wireless Sensor Networks (WSNs) has continued to encounter the problem of energy usage, network lifetime, and accurate data delivery, especially on non-homogeneous deployments on which the computational and energy capacities of the nodes are different. The old fashioned clustering protocols lack equal distribution and tend to choose cluster-heads, this is inefficient in the routing and has low adaptability to the dynamic nature of the network. In this paper, a new framework, named OEEMLCR (Optimized Energy-Efficient Machine Learning-Based Clustering and Routing), is proposed, integrating Particle Swarm Optimization-enhanced K -means clustering with Coati Optimization Algorithm-based routing and Q -learning adaptation. The suggested solution partially bridges the most problematic gap as it uses a two-fold objective fitness which aims to optimize both the spatial compactness and energy homogeneity during the formation of a cluster, and it uses reinforcement learning that allows adaptive routing behavior due to the gained network experience. Interaction-based experimental validation in heterogeneous network environments shows that OEEMLCR achieves significant gains over the current protocols: by 13.6\% network lifetime, 108\% cumulative packet delivery and 340\% energy efficiency. The framework also exhibits a constant within 1000 rounds of simulation and minimal power loss of 96.7\% of initial network energy, which was markedly superior in terms of lifetime, throughput, ratios of packet delivery and network resilience than LEACH, DMHT, EDMHT, and EEMLCR.