Adaptive intelligent resource optimization in wireless sensor networks using reinforcement learning and compressive sensing

Abstract

Abstract In response to the energy and memory limitations of WSNs, this research work introduces a new approach: EMPIRE-CS (Energy and Memory-optimized Performance with Iterative Reinforcement and Compressive Sensing). It is a hybrid approach combining compressive sensing with reinforcement learning and game-theoretic clustering. The architecture supports dynamic cluster head selection, fair energy distribution and effective signal compression with an iterative feedback mechanism between the sensor nodes and the sink. Compressive sensing decreases sampling and storage burdens, while reinforcement learning allows adaptive measurement rates and transmission scheduling. Energy-efficient roles of cluster in multi-round evolution are achieved by game-theoretic coalition strategies, which guarantee the fair allocation of energy-intensive responsibilities and the stability of cluster. The simulations on groups of nodes ranging from 200 to 1000 demonstrate that EMPIRE-CS is working. The method enhances network life time by 36.6%, retains up to 67.8% more residual energy, increases packet delivery ratio by 12.5%, and delivers up to 98% during dense deployment. Signal precision is improved, reconstruction error decreased by about 55.5% comparing with the baseline schemes. Memory efficiency is optimized, 14.2% compression efficiency gains reduce communication and storage costs. Cluster formation is also more robust with a stability gain of 150%, which avoids overhead for reconfiguration and energy wastage.