IoT-Enabled Logistics Optimization Framework for Real-Time Supply Chain Management

This paper introduces IoT-Enabled Logistics Optimization Framework, which is aimed at optimizing the supply chain management in the context of modern supply chain management. The suggested system combines Adaptive Kalman filtering to perform sensor-fusion smoothing of sensor readings, ReliefF feature selection to select the most significant logistics parameters and a Graph Attention Network classifier, which is realized within the FedML framework to perform federated and privacy-preserving learning. The combination of these technologies, the framework allows optimizing the paths dynamically, preventive maintenance, and resource use with lower latency and communication costs. The results of the experiment show that there is a dramatic increase in accuracy, reliability, and scalability, with a prediction accuracy of more than 94 percent and a significant decrease in latency and energy use. The offered solution contains a powerful and smart infrastructure of supply chain optimization in real-time, which allows to deal with the main challenges of data heterogeneity, system scalability and safe distributed learning of IoT-powered logistic ecosystems.