Machine and Deep Learning Driven Energy Efficient Clustering in IOT-WSNs: A Review

Energy-efficient clustering is a major challenge in IoT-enabled Wireless Sensor Networks (WSNs) because of limited battery power, node heterogeneity, and dynamic network topologies. This paper reviews recent developments in clustering protocols based on metaheuristics, machine learning (ML), deep learning (DL), hybrid optimization, trustaware security, and energy-harvesting strategies. A taxonomy of six categories is presented, along with a comparative evaluation across key metrics such as energy consumption, network lifetime, packet delivery ratio (PDR), latency, scalability, and throughput. The review shows that ML/DL and hybrid approaches usually perform better than traditional metaheuristics in terms of lifetime, adaptability, and reliability, though at the cost of higher computational complexity. However, issues such as static-topology assumptions, centralized bottlenecks, inference overhead, and limited realworld validation remain. To address these gaps, a roadmap is outlined focusing on lightweight model design, federated and distributed intelligence, integration of trust and security, energy harvesting, and cross-layer/SDN-based optimization. This work provides both a broad overview and practical insights to support the design of scalable and intelligent clustering solutions for future IoT-WSNs.