Fourier neural operator for learning solutions to macroscopic traffic flow models: Application to the forward and inverse problems

TL;DR

This paper introduces a physics-informed Fourier neural operator framework for accurately learning solutions to traffic flow models, effectively handling forward and inverse problems with sparse and heterogeneous data.

Contribution

It proposes a novel neural operator approach incorporating physics-based regularization to improve shock prediction and generalization in traffic flow modeling.

Findings

Superior accuracy in predicting traffic density dynamics.

Effective with simple training data for complex scenarios.

Sub-linear growth of extrapolation error with input complexity.

Abstract

Deep learning methods are emerging as popular computational tools for solving forward and inverse problems in traffic flow. In this paper, we study a neural operator framework for learning solutions to nonlinear hyperbolic partial differential equations with applications in macroscopic traffic flow models. In this framework, an operator is trained to map heterogeneous and sparse traffic input data to the complete macroscopic traffic state in a supervised learning setting. We chose a physics-informed Fourier neural operator ($\pi$-FNO) as the operator, where an additional physics loss based on a discrete conservation law regularizes the problem during training to improve the shock predictions. We also propose to use training data generated from random piecewise constant input data to systematically capture the shock and rarefied solutions. From experiments using the LWR traffic flow model, we found superior accuracy in predicting the density dynamics of a ring-road network and urban signalized road. We also found that the operator can be trained using simple traffic density dynamics, e.g., consisting of $2-3$ vehicle queues and $1-2$ traffic signal cycles, and it can predict density dynamics for heterogeneous vehicle queue distributions and multiple traffic signal cycles $(\geq 2)$ with an acceptable error. The extrapolation error grew sub-linearly with input complexity for a proper choice of the model architecture and training data. Adding a physics regularizer aided in learning long-term traffic density dynamics, especially for problems with periodic boundary data.