Learning Traffic Speed Dynamics from Visualizations

TL;DR

This paper introduces a deep learning approach to derive high-resolution traffic speed dynamics from space-time visualizations, enhancing traffic state estimation accuracy and robustness without relying on initial conditions or external factors.

Contribution

The method learns traffic speed dynamics directly from visualizations, offering finer resolution and causality-respecting estimations, outperforming existing approaches.

Findings

High-resolution traffic speed fields successfully estimated from datasets.

The approach improves robustness and independence from initial conditions.

Vehicle trajectories inferred from estimated speeds demonstrate practical utility.

Abstract

Space-time visualizations of macroscopic or microscopic traffic variables is a qualitative tool used by traffic engineers to understand and analyze different aspects of road traffic dynamics. We present a deep learning method to learn the macroscopic traffic speed dynamics from these space-time visualizations, and demonstrate its application in the framework of traffic state estimation. Compared to existing estimation approaches, our approach allows a finer estimation resolution, eliminates the dependence on the initial conditions, and is agnostic to external factors such as traffic demand, road inhomogeneities and driving behaviors. Our model respects causality in traffic dynamics, which improves the robustness of estimation. We present the high-resolution traffic speed fields estimated for several freeway sections using the data obtained from the Next Generation Simulation Program (NGSIM) and German Highway (HighD) datasets. We further demonstrate the quality and utility of the estimation by inferring vehicle trajectories from the estimated speed fields, and discuss the benefits of deep neural network models in approximating the traffic dynamics.