AAI-EAR: Adaptive AI-Based Energy-Aware Routing for Energy-Harvesting IoT Mesh Networks

Specifically, in the power-limited environment like energy harvesting-enabled IoT mesh networks face high energy consumption, poor resource management, and network strain-related challenges. AI can help in optimizing routing to improve energy efficiency and enhance the sustainability of nodes in IoT networks. In this paper, we proposed an adaptive AI-based energy-aware routing protocol (AAI-EAR) using the GNNs hybrid with Dijkstra's Algorithm. Further, we trained and evaluated the proposed AAI-EAR technique on a collected dataset from a hybrid star-mesh topology heterogenous IoT sensor's network. The GNN-based model initially predicts edge weights with an accuracy of 94.04% and forecasts battery level with a 0.77 RMSE score. Using these predictions, Dijkstra's Algorithm computes the energy-aware routes. Experimental results showed that the proposed AAI-EAR technique achieved 63.76% average energy efficiency in routing with a 97.20% successful packet delivery ratio.