A Continuous-time Link-based Kinematic Wave Model for Dynamic Traffic Networks

Ke Han, Benedetto Piccoli, Terry L. Friesz, Tao YaoarXiv 2012 · arXiv

Abstract

We present a continuous-time link-based kinematic wave model (LKWM) for dynamic traffic networks based on the scalar conservation law model. Derivation of the LKWM involves the variational principle for the Hamilton-Jacobi equation and junction models defined via the notions of demand and supply. We show that the proposed LKWM can be formulated as a system of differential algebraic equations (DAEs), which captures shock formation and propagation, as well as queue spillback. The DAE system, as we show in this paper, is the continuous-time counterpart of the link transmission model. In addition, we present a solution existence theory for the continuous-time network model and investigate continuous dependence of the solution on the initial data, a property known as well-posedness. We test the DAE system extensively on several small and large networks and demonstrate its numerical efficiency.

Introduction

Traffic flow modeling at the network level is a cornerstone of transportation engineering. Traditionally, the Lighthill-Whitham-Richards (LWR) model—a partial differential equation (PDE)—has been the standard for describing traffic density evolution. However, applying LWR to complex networks with intersections is mathematically challenging. Existing solutions often rely on discrete-time models like the Cell Transmission Model (CTM) or Link Transmission Model (LTM). While effective for simulation, these lack a rigorous continuous-time analytical framework. This paper bridges this gap by introducing a continuous-time link-based kinematic wave model (LKWM) formulated as a system of Differential Algebraic Equations (DAEs), capable of capturing critical phenomena like queue spillback with mathematical guarantees of well-posedness.

Methodology

The LKWM shifts from traditional density-based PDEs to a more manageable link-based formulation. A key innovation is the introduction of a new pair of state variables: flow ($q$) and regime ($r$).

Triangular fundamental diagram showing flow $q$ and regime $r$.

As shown in Figure 1, the regime $r$ is a binary variable where $r=0$ denotes free-flow and $r=1$ denotes congested traffic. This allows for a unique identification of the traffic state on a triangular fundamental diagram. The model utilizes the Lax-Hopf formula, a variational principle for the Hamilton-Jacobi equation, to solve for boundary conditions without simulating the entire interior of a link.

By integrating link dynamics with Riemann solvers—mathematical rules governing traffic behavior at merges and diverges—the authors derive a DAE system. This system represents the continuous-time equivalent of the LTM, providing a theoretical foundation for its accuracy and efficiency.

Results and Discussion

The proposed DAE system was rigorously tested for mathematical soundness and numerical performance. The authors established existence and uniqueness theorems for solutions at merge and diverge junctions, ensuring the model's stability for optimization tasks.

Graphical solution of the merge junction model under different demand/supply conditions.

Figure 6 illustrates the solution space for a merge junction. The shaded regions represent the feasible flows based on upstream demand ($D_4, D_5$) and downstream supply ($S_6$). The model identifies the optimal flow point (the black dot) that maximizes throughput while respecting physical constraints. This logic is central to capturing "spillback," where a bottleneck on a downstream link causes queues to propagate into upstream links.

Numerical experiments on synthetic networks demonstrated that the link-based approach is significantly more efficient than cell-based models, as it avoids the computational overhead of dividing roads into small segments and the associated numerical diffusion.

Conclusion

The LKWM provides a mathematically rigorous and computationally efficient framework for network-level traffic modeling. By formulating traffic dynamics as a DAE system, it offers a continuous-time counterpart to popular discrete models like LTM. While the current model assumes a triangular fundamental diagram and fixed turning ratios, it sets the stage for future work in Dynamic User Equilibrium (DUE) and large-scale network optimization.

Summary updated

Abstract

We present a continuous-time link-based kinematic wave model (LKWM) for dynamic traffic networks based on the scalar conservation law model. Derivation of the LKWM involves the variational principle for the Hamilton-Jacobi equation and junction models defined via the notions of demand and supply. We show that the proposed LKWM can be formulated as a system of differential algebraic equations (DAEs), which captures shock formation and propagation, as well as queue spillback. The DAE system, as we show in this paper, is the continuous-time counterpart of the link transmission model. In addition, we present a solution existence theory for the continuous-time network model and investigate continuous dependence of the solution on the initial data, a property known as well-posedness. We test the DAE system extensively on several small and large networks and demonstrate its numerical efficiency.
BibTeX
@misc{han2012continuous,
  author = {Ke Han and Benedetto Piccoli and Terry L. Friesz and Tao Yao},
  title = {{A Continuous-time Link-based Kinematic Wave Model for Dynamic Traffic Networks}},
  year = {2012},
  eprint = {1208.5141},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1208.5141},
}

Loading related content...

Figure