In classical fluid dynamics and macroscopic traffic modeling, systems are governed by hyperbolic conservation laws. These partial differential equations (PDEs) describe how a continuous quantity, like traffic density, evolves over time. When traffic hits a bottleneck—such as a lane reduction—vehicles bunch up, creating a sharp jump in density known as a shock wave. Standard numerical solvers handle these situations well by enforcing mathematical "entropy conditions," which ensure that these density jumps are physically realistic and stable.
However, classical models break down when the bottleneck itself moves. Consider a slow-moving bus or an autonomous vehicle (AV) traveling down a single-lane highway. This vehicle acts as a localized capacity restriction that moves through the spatial domain over time. Modeling this interaction requires coupling the continuous density field of the traffic (a PDE) with the discrete trajectory of the moving obstacle (an ordinary differential equation, or ODE).
This coupling introduces a moving, unilateral inequality constraint on the system's flux function. Because the constraint is driven by an external, mobile obstacle rather than the internal dynamics of the fluid, it generates non-classical shock waves—discontinuities that violate standard entropy conditions and baffle traditional numerical solvers.
Why Moving Bottlenecks Defy Classical Theory
To understand why moving bottlenecks require a fundamentally different mathematical approach, we must look at how density behaves right at the obstacle's location.
In a standard traffic jam caused by a stationary red light, density builds up upstream and drops downstream. When the light turns green, vehicles gradually accelerate, creating a smooth transition called a rarefaction wave.
A moving bottleneck behaves differently. As explored by Maria Laura Delle Monache in Modeling of moving bottlenecks in traffic flow: a PDE-ODE model with moving density constraints, the presence of a slow-moving vehicle forces an abrupt, localized drop in capacity. Traffic accumulates immediately behind the obstacle, but the moment vehicles pass it, they instantly accelerate into the clearer road ahead.
This creates a sharp, persistent jump discontinuity where density drops abruptly across the moving bottleneck. In classical conservation laws, such an immediate drop in density across a shock is mathematically forbidden by standard entropy conditions; traditional solvers view it as an unphysical "rarefaction shock" and will incorrectly smooth it out. Yet, in the presence of a moving constraint, this non-classical shock is the correct physical reality.
Classical Shock (Stationary Bottleneck)
Density (ρ)
▲
│ ┌───────────
│ │
│─────┘
└──────────────────► Space (x)
▲
Bottleneck
Non-Classical Shock (Moving Bottleneck)
Density (ρ)
▲
│ ───┐ ◄── High density behind moving obstacle
│ │
│ └─────────── ◄── Immediate acceleration/low density ahead
└──────────────────► Space (x)
▲
Moving Obstacle y(t)
The PDE-ODE Coupling Mechanism
To rigorously model this phenomenon, researchers frame the problem as a strongly coupled PDE-ODE system. As outlined by Paola Goatin in Scalar conservation laws with moving constraints arising in traffic flow modeling: an existence result, the traffic density $\rho(x,t)$ is governed by a conservation law, while the bottleneck's position $y(t)$ evolves based on its own desired speed and the surrounding traffic density:
$$\partial_t \rho + \partial_x f(\rho) = 0$$
$$\dot{y}(t) = v(t)$$
The core innovation lies in how these two equations interact. The moving bottleneck imposes a strict inequality constraint on the flux function $f(\rho)$ at the exact coordinate of the obstacle:
$$f(\rho(y(t), t)) - \dot{y}(t)\rho(y(t), t) \le F_\max(\dot{y}(t))$$
This inequality states that the flow of traffic relative to the moving bottleneck cannot exceed a maximum capacity $F_\max$, which depends on the bottleneck's velocity. If the surrounding traffic density is low, the constraint is inactive, and the vehicle moves at its free-flow speed. If traffic is dense, the constraint becomes active, forcing a non-classical jump discontinuity in the density field at $x = y(t)$.
This framework is highly versatile. For instance, Paola Goatin et al. extended this concept in Interacting moving bottlenecks in traffic flow to simulate how multiple autonomous vehicles act as dynamic, interacting flux constraints, altering macroscopic traffic patterns and generating complex non-classical wave patterns.
Resolving Non-Classical Shocks Numerically
Standard finite volume methods (like the Godunov or Lax-Friedrichs schemes) are optimized for classical shocks. When applied to moving constraints, they suffer from severe numerical smearing or fail to capture the capacity drop entirely because they lack a mechanism to enforce localized, moving inequalities.
To overcome this, specialized numerical frameworks must be introduced. One approach to handling abrupt changes in flow properties involves domain decomposition. Mohammed Seaı̈d et al. demonstrated in A domain decomposition method for conservation laws with discontinuous flux function that breaking a complex conservation law with a discontinuous flux into simpler sub-problems with continuous fluxes can stabilize numerical solutions at interfaces.
For moving constraints, however, the interface travels across the computational grid over time. In A numerical scheme for moving bottlenecks in traffic flow, Paola Goatin introduced a specialized Riemann solver that explicitly embeds the local flux constraint into the cell interface calculations.
When a bottleneck sits inside a computational cell, the scheme dynamically reconstructs the states on either side of the obstacle. It calculates the exact time required for these reconstructed discontinuities to reach the cell interfaces, adjusting the numerical fluxes accordingly.
Standard Scheme (Smeared Interface)
Cell i-1 Cell i Cell i+1
┌───────────┬───────────┬───────────┐
│ Density │ Smeared │ Density │
│ High │ Gradient │ Low │
└───────────┴───────────┴───────────┘
▲
Moving Obstacle
Specialized Scheme (Sharp Interface Tracking)
Cell i-1 Cell i Cell i+1
┌───────────┬─────┬─────┬───────────┐
│ Density │High │ Low │ Density │
│ High │ │ │ Low │
└───────────┴─────┴─────┴───────────┘
▲
Explicit Boundary Condition
Building on this, Christophe Chalons et al. developed a fully conservative scheme in A conservative scheme for non-classical solutions to a strongly coupled PDE-ODE problem. Their method enforces specific, sharp boundary conditions at the moving obstacle's precise location. By treating the bottleneck as a sharp, moving interface rather than a smeared volumetric change, the scheme prevents numerical diffusion. This allows the solver to remain stable and perfectly resolve the forced, non-classical density drops that define real-world moving bottlenecks.
Key Foundational Work
Modeling of moving bottlenecks in traffic flow: a PDE-ODE model with moving density constraints (2014) — contributed 4 concept elements to this direction.
A numerical scheme for moving bottlenecks in traffic flow (2014) — contributed 4 concept elements to this direction.
Scalar conservation laws with moving constraints arising in traffic flow modeling: an existence result (2014) — contributed 4 concept elements to this direction.