Traffic Prediction and Congestion Control Using an Enhanced-Graph Neural Network

The increasing number of automobiles on the highway has led to a major difficulty in municipal traffic management.Intelligent Transportation Systems (ITS) require dependable traffic prediction algorithms capable of providing accurate forecasts at numerous time steps.This research proposes an Enhanced-Graph Neural Network (E-GNN) technique for traffic prediction and has been explored to augment the traditional GNN and temporal dependencies in traffic networks.A multimodal input was deployed for the preprocessing of the input data with GNN-Layer.An additional data stream was integrated to influence the traffic flow.The approach leverages strategically positioned loop detector sensors on the road network as a means of harvesting real-world traffic data.The suggested E-GNN technique for the estimation of real-time traffic speed was developed using two separate actual traffic datasets, such as PeMS-BAY and METR-LA.The result obtained over time shows a significant improvement, as seen in the 15-minute ahead prediction; the RMSE of EGNN reduced by 26.25% when compared with the existing state-of-the-art techniques.