An RL-Based SWIPT Optimization Framework for Energy-Autonomous MIMO Wireless Sensor Networks

Abstract

Wireless sensor networks (WSNs) are critical for Industrial Internet of Things (IoT), healthcare, and environmental monitoring, yet limited energy resources constrain their reliability and longevity. This article presents a hybrid reinforcement learning (RL) framework to optimize simultaneous wireless information and power transfer (SWIPT) in multiple-input-multiple-output (MIMO)-enabled WSNs, enabling energy-autonomous and sustainable sensing. By integrating sequential parametric convex approximation (SPCA) with state-action-reward-state-action (SARSA) and Q-learning, the framework employs power-splitting (PS) and time-switching (TS) techniques to enhance routing efficiency and energy harvesting (EH) in dynamic sensor fields. A nonlinear EH model captures practical circuit constraints, such as diode sensitivity and leakage currents, improving prediction accuracy. Simulations in a 1000 & times; 1000 m(2) area with distributed sensor nodes show up to 20 improvement in energy efficiency and 15 increase in data throughput over baseline methods. These advancements position the framework as a transformative solution for energy-constrained WSNs in 5G/6G-enabled IoT and smart sensor applications, paving the way for sustainable large-scale deployments.