Energy-Aware Reinforcement-Driven Clustering Framework with Adaptive Processing for Sustainable IoT Sensor Networks

Abstract

Energy saving and sustainable behavior of the system is a vital problem, especially for a resource-limited agricultural monitoring environment in WSNs. In this paper, we introduce a Reinforcement-driven Energy-aware Clustering scheme with Adaptive Processing (RECAP) for extending lifetime, conserving energy, and maintaining data fidelity in field situations. RECAP adopts a hierarchical architecture comprising an Adaptive Duty-Cycling Engine (ADCE), an Energy-Aware Clustering Module (EACM), a Reinforcement Learning Controller (RLC), and an Edge Processing and Communication Layer (EPCL). Every Cluster Head tunes their work dynamically by a lightweight Q-learning model that adapts cluster formation, transmission scheduling, and node sleeping cycles according to local energy and topology states. The architecture uses a hybrid energy draining model through context-aware duty cycling and data aggregation to avoid redundant transmissions. Simulation results show that RECAP reduces 68% energy consumption, achieves event-triggered data collection with 98% accuracy and low computation overhead, and overcomes the latency problem than existing hybrid models like LEACH-CNN, PSO-LSTM, and base GA-GRU. These results validate RECAP for sustainable IoT-based precision agriculture and other mission-critical WSN deployments.