Exploring DMD-type Algorithms for Modeling Signalised Intersections

TL;DR

This paper applies Koopman operator theory and dynamic mode decomposition to model complex traffic dynamics at signalized intersections, enabling linear approximation for better prediction and control.

Contribution

It introduces a data-driven Koopman-based DMD approach for modeling nonlinear traffic dynamics at intersections, enhancing prediction accuracy over traditional methods.

Findings

DMD algorithms effectively capture complex traffic dynamics.

The approach outperforms LSTM in queue length prediction.

Linear models provide useful insights for adaptive traffic control.

Abstract

This paper explores a novel data-driven approach based on recent developments in Koopman operator theory and dynamic mode decomposition (DMD) for modeling signalized intersections. Vehicular flow and queue formation on signalized intersections have complex nonlinear dynamics, making system identification, modeling, and controller design tasks challenging. We employ a Koopman theoretic approach to transform the original nonlinear dynamics into locally linear infinite-dimensional dynamics. The data-driven approach relies entirely on spatio-temporal snapshots of the traffic data. We investigate several key aspects of the approach and provide insights into the usage of DMD-type algorithms for application in adaptive signalized intersections. To demonstrate the utility of the obtained linearized dynamics, we perform prediction of the queue lengths at the intersection; and compare the results with the state-of-the-art long short term memory (LSTM) method. The case study involves the morning peak vehicle movements and queue lengths at two Orlando area signalized intersections. It is observed that DMD-based algorithms are able to capture complex dynamics with a linear approximation to a reasonable extent.