Physics-informed Machine Learning for Calibrating Macroscopic Traffic Flow Models

TL;DR

This paper introduces a physics-informed machine learning method combining autoencoders with traffic flow models to improve calibration accuracy, handle noisy data, and outperform traditional optimization techniques.

Contribution

It presents a novel autoencoder-based calibration approach that integrates physical traffic models and manages corrupted data, enhancing calibration performance.

Findings

Achieves comparable or better calibration accuracy than optimization methods.

Effectively handles noisy and missing data in traffic measurements.

Validated with a case study on California's I-210 E.

Abstract

Well-calibrated traffic flow models are fundamental to understanding traffic phenomena and designing control strategies. Traditional calibration has been developed base on optimization methods. In this paper, we propose a novel physics-informed, learning-based calibration approach that achieves performances comparable to and even better than those of optimization-based methods. To this end, we combine the classical deep autoencoder, an unsupervised machine learning model consisting of one encoder and one decoder, with traffic flow models. Our approach informs the decoder of the physical traffic flow models and thus induces the encoder to yield reasonable traffic parameters given flow and speed measurements. We also introduce the denoising autoencoder into our method so that it can handles not only with normal data but also with corrupted data with missing values. We verified our approach with a case study of I-210 E in California.