Modeling of spacing distribution of queuing vehicles at signalized junctions using random-matrix theory

Xuexiang Jin, Yuelong Su, Yi Zhang, Zheng Wei, Li LiTsinghua 2009 · DOI · arXiv

Abstract

The modeling of headway/spacing between two consecutive vehicles in a queue has many applications in traffic flow theory and transport practice. Most known approaches have only studied vehicles on freeways. This paper presents a model for the spacing distribution of queuing vehicles at a signalized junction based on random-matrix theory. The spacing distribution of a Gaussian symplectic ensemble (GSE) fits well with recently measured spacing distribution data. These results are also compared with measured spacing distribution observed for the car parking problem. Vehicle stationary queuing and vehicle parking have different spacing distributions due to different driving patterns.

Introduction

While traffic flow on open freeways is a well-established field of study, the "static" phase—where vehicles are stopped in a queue at signalized junctions—receives significantly less attention. The spacing between these vehicles is not merely random; it reflects a complex psychological and physical balance between a driver's desire to minimize gaps for efficiency and the necessity of maintaining a safe distance. This paper addresses the need for a universal mathematical framework to describe these spacing distributions, investigating whether the interaction potentials between human drivers mirror those found in complex physical systems.

Theoretical Framework: The Coulomb Gas Analogy

To model this behavior, the researchers employed Dyson’s Coulomb gas model from Random-Matrix Theory (RMT). In this framework, queuing vehicles are treated as point charges on a one-dimensional line. The system's potential energy is defined by two competing factors: a harmonic potential that draws vehicles toward the intersection (representing the collective desire to move forward) and an electrostatic-like repulsion between adjacent vehicles (representing the safety buffer).

Schematic representation of queuing vehicles and the interaction forces modeled using Random-Matrix Theory.

The study tested empirical data against four standard RMT distributions, characterized by the parameter $\beta$, which represents the strength of repulsion:

  • Poisson ($\beta=0$): Represents a lack of interaction (random spacing).
  • GOE ($\beta=1$): Gaussian Orthogonal Ensemble.
  • GUE ($\beta=2$): Gaussian Unitary Ensemble, previously associated with car parking spacing.
  • GSE ($\beta=4$): Gaussian Symplectic Ensemble, representing a higher degree of order.

Empirical Analysis and Results

The model was validated using a dataset of 700 sample spacings collected from signalized junctions in Beijing, China. The data was normalized by the mean observed spacing (1.43 meters) to facilitate comparison with theoretical RMT curves.

Comparison of empirical spacing data from Beijing with theoretical RMT distributions, showing the superior fit of the GSE model.

As illustrated in the comparison above, the normalized histogram of the Beijing traffic data aligns remarkably well with the GSE ($\beta=4$) theoretical curve. This distribution significantly outperforms the Poisson, GOE, and GUE models in capturing the statistical fluctuations of the queue.

Discussion: "Rigid" vs. "Loose" Constraints

The finding that queuing vehicles follow the GSE distribution ($\beta=4$) rather than the GUE distribution ($\beta=2$)—which fits parallel parking—is significant. The authors attribute this difference to the "degrees of freedom" inherent in different driving patterns.

In a parking scenario, drivers can move both forward and backward to adjust their position, resulting in "loose" repulsion. Conversely, in a traffic queue at a signal, drivers typically only move forward and stop. The inability to easily reverse and adjust the gap creates a "rigid" constraint, leading to the higher repulsion strength characterized by $\beta=4$. This mathematically captures the human instinct to find a "sweet spot" that avoids both excessive gaps and unsafe proximity.

Conclusion

This research demonstrates that human-driven systems can exhibit the same statistical signatures as complex physical systems, such as heavy-atom nuclei. The identification of the GSE distribution as a fit for queuing traffic suggests that "RMT-like fluctuations" are a universal feature of constrained multi-agent systems.

While the study is currently limited by its geographic focus on Beijing, it provides a robust foundation for future work. Variations in mean spacing across different global cities may exist due to cultural factors or vehicle sizes, but the underlying GSE distribution is expected to hold as a universal descriptor of queuing dynamics.

Summary updated

Abstract

The modeling of headway/spacing between two consecutive vehicles in a queue has many applications in traffic flow theory and transport practice. Most known approaches have only studied vehicles on freeways. This paper presents a model for the spacing distribution of queuing vehicles at a signalized junction based on random-matrix theory. The spacing distribution of a Gaussian symplectic ensemble (GSE) fits well with recently measured spacing distribution data. These results are also compared with measured spacing distribution observed for the car parking problem. Vehicle stationary queuing and vehicle parking have different spacing distributions due to different driving patterns.
BibTeX
@inproceedings{jin2009modeling,
  author = {Xuexiang Jin and Yuelong Su and Yi Zhang and Zheng Wei and Li Li},
  title = {{Modeling of spacing distribution of queuing vehicles at signalized junctions using random-matrix theory}},
  booktitle = {Tsinghua},
  year = {2009},
  doi = {10.1016/s1007-0214(09)70037-9},
}

Loading related content...

Figure