EDEN: Energy-aware Dynamic Genetic and Neural Network-based Path Predictive Routing and Clustering for Mobile SD-IoT Networks
Negar Javadzadeh No, Hossein Taghizadeh, Mohammad Parsa Sedighi, Bardia Safaei, Jörg Henkel · 2025
The proliferation of battery-equipped smart devices in Internet of Things (IoT) applications has underscored the critical need for energy-efficient communication and computation solutions to extend lifetime. Meanwhile, the complex nature of mobile IoT networks’ communications poses significant challenges. Software-Defined Networking (SDN) offers minimized device-related overheads associated with processing and computations by centralizing energy-intensive tasks. The employed central controller in SDN can effectively follow, and manage the continuous topological alterations in dynamic mobile environments, thereby mitigating energy overhead imposed on individual IoT devices. On the other hand, clustering can further improve energy efficiency in IoT networks by reducing the number of transmissions, aggregating data efficiently, balancing the load among nodes, and optimizing routing paths. Accordingly, this paper introduces EDEN; an energy-aware SDN-based routing and clustering approach for mobile IoT networks to reduce energy consumption, increase network lifetime, and enhance reliability in the network in terms of Packet Delivery Ratio (PDR). To determine the optimal number of clusters and ensure a balanced distribution, EDEN utilizes a dynamic genetic algorithm to adaptively determine mutation and crossover rates. In selecting cluster heads, EDEN incorporates multiple objective function parameters, including node centrality, remaining energy, and distance, to optimize energy efficiency. Furthermore, EDEN employs a path prediction algorithm based on the LSTM Neural Network (NN) to forecast the trajectory of the mobile nodes to maintain cluster stability and reduce the frequency of re-clustering, thereby enhancing both energy efficiency and reliability. Extensive simulations in the NS3 environment demonstrate the effectiveness of the proposed solution, showing improvements in energy consumption by at least 32% while improving PDR by more than 98% compared to the state-of-the-art.