Identification of Parameters in a Freeway Traffic Model

TLDR

Macroscopic models represent traffic flow in terms of volume, density, and speed. The study explores extending the methodology to use real freeway traffic data from automated surveillance systems. The authors applied a discrete‑time extended Kalman filter to identify parameters of a macroscopic freeway traffic model, first verifying local identifiability at nominal values, and used data generated from a microscopic simulation of individual vehicle movements. The study demonstrates local identifiability of the model and successfully identifies two key parameters—reaction time and density‑sensitivity—using the proposed methodology.

Abstract

The methodology of discrete time, extended Kalman filtering is applied to the problem of identifying parameters of a macroscopic freeway traffic model. Macroscopic models provide a representation of traffic flow in terms of its gross properties, i.e., volume, density, and speed. The local identifiability of a parameterization of macroscopic model at nominal values of the unknown parameters is checked before any identification is attempted. It is shown that the parameterization is locally identifiable. Two parameters of the model (reaction time and sensitivity to changing density) were identified through the use of this methodology. The data base for studies to date was generated from a microscopic simulation of freeway traffic, which involves following all individual vehicle movements. Techniques for extending the methodology to employ real freeway traffic data, especially as can be obtained from automated surveillance systems, are discussed.