Hybrid Energy-Efficient Clustering With Reinforcement Learning for IoT-WSNs Using Knapsack and K -Means

Wireless sensor networks (WSNs) play a fundamental role in the Internet of Things (IoTs), with widespread applications in areas such as smart city infrastructure, industrial control systems, and environmental monitoring. Despite their broad utility, challenges related to energy efficiency and network longevity persist. To address these issues, this article introduces an integrated framework that leverages distributed energy-efficient clustering (DEEC), the energy-efficient Knapsack algorithm (EEKA), K-means clustering, and reinforcement learning (RL) to optimize energy usage and improve overall network performance. While cluster-heads (CHs) selection is conducted in a balanced manner by DEEC to prevent energy depletion, the selection of sensor node CHs also avoids exhausting sensor node energy. EEKA optimizes task allocation under energy constraints, allowing for efficient distribution of tasks based on available energy levels. This approach conserves more energy spent on data transmission compared to other techniques, as K-means clustering minimizes intracluster communication overhead. Moreover, during varying conditions such as cluster sizes or transmission power, parameters related to the network can be adaptively tuned in real-time by RL, enhancing stability and performance. Simulation results demonstrate a significant improvement in energy efficiency with this hybrid model compared to traditional approaches under similar conditions. Therefore, for sustainable energy management within IoT-enabled WSNs, the DEEC-EEKA–K-means-RL framework offers a robust adaptive method for efficient resource utilization across diverse operational contexts, rather than relying solely on robustness and resilience-based solutions.