Energy-Efficient Real-Time Localization for Distributed Iot Sensing: Novel CNN-Based Models

The rapid expansion of the Internet of Things (IoT) in sectors such as agriculture, healthcare, smart cities, and industrial automation has increased demand for low-energy, highprecision devices for real-time localization. However, accurate localization in large-scale, resource-constrained IoT networks remains challenging due to increased data communication, energy consumption, and network dynamics. Traditional methods such as GPS, triangulation, and RSSI often fail in obstructed or indoor environments. Although machine learning techniques such as Convolutional Neural Networks (CNNs) offer potential, their data and computational requirements limit their use in IoT settings. Similarly, stochastic models and swarm intelligence algorithms face issues such as slow convergence and difficulty modeling complex environments. This paper proposes hybrid models that integrate CNNs, stochastic models, and swarm intelligence to tackle localization challenges in dynamic IoT networks. We introduce two novel energy-efficient hybrid frameworks, MarkovCNN and SSA-Markov-CNN, which combine CNNs for feature extraction, Markov models for uncertainty management, and swarm intelligence for global optimization. Experiments show that these models significantly improve localization accuracy while reducing energy consumption and latency, offering a promising solution for energy-conscious IoT deployments. Our findings provide insights for selecting optimal localization models for low-latency, energy-efficient IoT applications.