International Journal of Scientific Research and Engineering Development

Traffic congestion in urban environments poses a significant challenge due to increasing vehicle density and complex traffic dynamics. Traditional traffic prediction models, including statistical and deep learning approaches, often fail to capture nonlinear traffic behavior and require large volumes of data. This paper proposes a Physics-Informed Neural Network (PINN) approach for traffic flow prediction based on the one-dimensional viscous Burgers’ Equation. The proposed model integrates deep learning with physical laws by embedding the governing partial differential equation into the loss function, ensuring both accuracy and physical consistency. The model is trained using a hybrid strategy that combines sparse data with physics-based collocation points. A composite loss function incorporating data loss, physics loss, and boundary constraints is optimized using Adam and L-BFGS optimizers.Experimental results demonstrate that the model achieves high prediction accuracy with a relative L2 error of approximately 0.42%, while requiring significantly less training data compared to conventional approaches. The model effectively captures nonlinear traffic phenomena such as shock waves and shows strong generalization capability. The proposed approach highlights the potential of integrating physics-based modeling with deep learning for efficient and reliable traffic prediction in intelligent transportation systems.

Mr.M V Ramana Murthy, Leela Krishna D, Nikhila B, Neeraj G, Vamsi G,"A Physics-Informed Neural Network Approach for Traffic Flow Prediction Using Burgers’ Equation" International Journal of Scientific Research and Engineering Development, V9(2): Page(1762-1765) Mar-Apr 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.