AIEH-CF: A Stability-Guaranteed Multi-Objective Energy Harvesting Framework for Ultra-Low-Power IoT Networks

One of the primary challenges for ultra-low power Internet of Things (IoT) networks is the availability of energy, especially if the IoT device needs to rely on an energy source that can only occasionally be found in its environment. Changes in harvested energy may cause miscommunication, uneven energy use among nodes, and reduced lifetime of network operation. To overcome these challenges, this study introduces AIEH-CF, an adaptive energy harvesting framework that aims to coordinate energy management and routing decisions in the presence of uncertainty in the operating conditions. The framework is designed to combine a Long Short-Term Memory (LSTM) network to predict energy availability, reinforcement learning (RL) to make routing decisions, and graph neural networks (GNNs) to optimize network topology and balance energy consumption. The routing problem is modelled as a constrained Markov decision process (CMDP) and Lyapunov drift-plus-penalty control is used as a mathematical tool to analyse the stability of the system. The proposed approach also considers solar and radio-frequency energy harvesters and real-world hardware limitations such as battery size, environmental variations, and communication needs. The simulation experiments under various network configurations reveal that the network lifetime is improved by up to 35.8% and energy efficiency is improved by up to 29.4% over existing energy-harvesting-aware routing protocols. Statistically it is also shown that the performances remain constant in various harvestings. The results show that there is a great potential for using coordinated prediction, routing and stability control to enhance the reliability and sustainability of energy-constrained IoT deployments.