Traffic Congestion Monitoring Improvement Through Federated Learning Technique

Abstract

The Internet of Things (IoT) has significantly influenced the management of intelligent traffic systems, mainly by incorporating Machine Learning (ML) for congestion detection. This study underscores the application of Federated Learning (FL) in identifying traffic congestion based on various factors, including stringent delay constraints and real-time speed data obtained from Differential GPS (DGPS). Specifically, the FedAvg (Federated Averaging) model within FL is utilized to forecast traffic speeds, incorporating diverse parameters such as training sets, prediction sets, and road sector data frames. The results are being validated by empowering the Explained Artificial Intelligence (EAI) approach, which employs a dynamic weight factor to detect congestion based on vehicle speeds. In conclusion, the model's outcomes highlight its proficiency, efficiency, and accuracy in managing traffic congestion effectively. The model significantly improves, demonstrating at least 10 to 15% greater efficiency than other approaches.