Multivalued Fundamental Diagrams of Traffic Flow in the Kinetic Fokker--Planck Limit

Giuseppe Visconti, Michaël Herty, Gabriella Puppo, Andrea Tosin

Introduction

In traffic engineering, the "Fundamental Diagram" is a cornerstone concept that relates vehicle density (the number of vehicles per unit length) to flux (the number of vehicles passing a point per unit time). While classical models often assume a simple one-to-one relationship between these variables, empirical sensor data consistently reveals a more complex reality. In free-flowing traffic, the relationship is predictable; however, as density increases, the data becomes "multivalued" or scattered. This research bridges the gap between microscopic driver behavior and these macroscopic observations by deriving a new mathematical model based on kinetic theory.

The Problem: Beyond Single-Curve Models

Traditional macroscopic models struggle to explain why, at the same traffic density, one might observe vastly different flow rates. This "scatter" in congested traffic is often dismissed as measurement noise or handled through high-complexity simulations. The authors of this paper argue that this behavior is not an anomaly but a fundamental result of microscopic stochasticity—the inherent uncertainty and varied decision-making of individual drivers. By starting at the mesoscopic (kinetic) scale, they aim to derive macroscopic patterns from the ground up.

Methodology: From Boltzmann to Fokker–Planck

The researchers utilize Kinetic Theory, a framework typically used in statistical mechanics to describe gas molecule interactions, and adapt it for vehicular flow. The process follows a rigorous mathematical pipeline:

  1. Microscopic Interaction Rules: The model defines how a driver adjusts their speed based on the mean speed of surrounding traffic. It incorporates two levels of stochasticity: the probability of accelerating versus braking based on local density, and a "noise" term representing human imperfection in reaching a target speed.
  2. The Grazing Collision Limit: Boltzmann equations, while accurate, are mathematically intractable due to complex interaction integrals. The authors simplify this by applying a "grazing collision limit" ((\epsilon \to 0)). This assumes that while vehicle interactions are frequent, each individual interaction results in only a minute change in speed.
  3. Stationary Distributions: This transformation yields a Fokker–Planck equation, a partial differential equation that is significantly easier to analyze. From this, the authors derive analytical expressions for how speeds are distributed at equilibrium.
  4. The $r$ Parameter: To account for the diversity of driver behavior, the model introduces a parameter ($r$) representing the ratio of "slow" to "fast" vehicles at a given mean speed. This allows the model to produce a family of stationary distributions rather than a single fixed state.

Key Insights and Empirical Validation

The model was validated against real-world data from two distinct environments: the Minnesota Highway system (representing typical US highway flow) and urban data from Rome, Italy (representing dense, chaotic urban traffic).

Stochasticity as a Feature

A primary insight of the paper is that the "scatter" seen in traffic data is a natural consequence of driver uncertainty. The model demonstrates that the famous Greenshields model—a linear relationship between speed and density—is actually a special case of this framework that occurs only when driver uncertainty drops to zero.

The Power of the $r$ Parameter

Instead of requiring separate equations for different vehicle types (e.g., trucks vs. cars), the $r$ parameter synthesizes heterogeneous traffic compositions into a single framework. This parameter effectively explains the multivalued nature of the flux-density diagram, as different values of $r$ correspond to different flow outcomes at the same density.

Capacity Drop and Phase Transitions

The model successfully reproduces the "capacity drop," a phenomenon where the maximum flow of a road suddenly decreases once a critical density is reached. It identifies the specific phase transitions where traffic shifts from a free-flowing state to a synchronized or congested state.

Conclusion and Future Directions

By providing explicit analytical formulas for vehicle speed distributions, this research offers a significant advancement over models that rely solely on numerical simulation. While the current model assumes a spatially homogeneous environment (a uniform road without ramps), it provides a robust foundation for future multi-population models. Future work will look to incorporate lane-changing dynamics and non-homogeneous spaces to better understand how traffic jams propagate across complex road networks.

Summary updated

Abstract

Starting from interaction rules based on two levels of stochasticity we study the influence of the microscopic dynamics on the macroscopic properties of vehicular flow. In particular, we study the qualitative structure of the resulting flux-density and speed-density diagrams for different choices of the desired speeds. We are able to recover multivalued diagrams as a result of the existence of a one-parameter family of stationary distributions, whose expression is analytically found by means of a Fokker--Planck approximation of the initial Boltzmann-type model.
BibTeX
@inproceedings{visconti2017multivalued,
  author = {Giuseppe Visconti and Michaël Herty and Gabriella Puppo and Andrea Tosin},
  title = {{Multivalued Fundamental Diagrams of Traffic Flow in the Kinetic Fokker--Planck Limit}},
  booktitle = {Multiscale},
  year = {2017},
  doi = {10.1137/16m1087035},
}
Figure