What question did this study set out to answer?

The study aims to develop an efficient framework for predicting traffic congestion using federated learning.

June 12, 2026

Intelligent Real‐Time Traffic Prediction in VANETs Using Federated Ensemble Neural Networks

Key Points

The study aims to develop an efficient framework for predicting traffic congestion using federated learning.
Proposed a Federated Ensemble Neural Network incorporating CNN, ResDense, and MLP architectures.
Employed an adaptive soft-voting mechanism for combining predictions.
Simulated using a dataset containing complex traffic patterns.
Achieved 94.27% accuracy in traffic predictions.
Obtained an R² score of 0.82 for capturing dense traffic dynamics.

Abstract

ABSTRACT Traffic congestion creates a significant challenge in urban environments that highly impacts infrastructure efficiency. Exact traffic prediction is important to avoid such challenges in difficult traffic scenarios. In this study, we propose a Federated Ensemble Neural Network (FENN) framework that integrates a lightweight Convolutional Neural Network (CNN), a Residual Dense (ResDense) network, and a Multi‐Layer Perceptron (MLP) within a distributed federated learning architecture with an adaptive soft‐voting mechanism. The suggested model predicts traffic congestion by extracting temporal features with a CNN and ResDense to model congestion patterns, capturing complex nonlinear traffic patterns with an MLP, and combining the predictions through soft voting to improve prediction efficiency. The model leverages knowledge from FedAvg‐based training to ensure consistent performance across multiple RSU clients. The model is simulated using a traffic prediction dataset with complex traffic patterns. Experimental results proved that the model achieves 94.27% accuracy and an R 2 score of 0.82 in capturing dense traffic dynamics. As a result, the model provides a promising solution for traffic prediction and congestion reduction in urban platforms.

AI से पूछें

Bookmark