On Hamilton Jacobi equations with time measurable Hamiltonians posed on a 1-dimensional junction

This paper introduces a flux-limited viscosity solution framework for evolutive Hamilton-Jacobi equations with time-measurable Hamiltonians and discontinuous flux limiters on a 1-dimensional junction, proving a comparison principle and establishing existence results for convex cases via optimal control.

The Gist

Imagine you are managing traffic on a very simple road network: two highways meeting at a single intersection (a "junction"). One highway goes to the right (Edge 1), and the other goes to the left (Edge 2).

Your goal is to predict how a wave of cars (or information, or heat) will move through this system over time. In the world of mathematics, this is described by an equation called the Hamilton-Jacobi equation.

Usually, mathematicians assume that the rules of the road are smooth and predictable. For example, the speed limit changes gradually, or the traffic lights follow a perfect, smooth rhythm.

The Problem:
In this paper, the author, Ariela Briani, tackles a much messier, more realistic scenario. She asks: What happens if the rules of the road are chaotic and unpredictable?

1. Time is messy: The speed limits or traffic light patterns don't change smoothly; they jump around randomly. Mathematically, they are only "measurable" (we can measure them, but we can't predict them smoothly).
2. The Junction is messy: The intersection itself has a "flux limiter." Think of this as a traffic cop or a bottleneck at the junction. This cop might be shouting orders, changing their mind instantly, or even taking a coffee break. Their behavior is also "messy" (only measurable, not smooth).

The Challenge:
Standard math tools for solving these traffic problems rely on smoothness. If you try to use a smooth tool on a jagged, chaotic problem, the tool breaks. The math gets stuck because you can't take the usual "derivatives" (rates of change) when things jump around.

The Solution: A New Kind of "Viscosity"
The paper introduces a new way to define a "solution" to this problem. The author calls it a "Flux-Limited t-Measurable Viscosity Solution." That's a mouthful, so let's break it down with an analogy:

• Viscosity Solution: Imagine you are trying to find the highest point on a bumpy, foggy landscape. You can't see the whole mountain. Instead, you place a smooth, flexible sheet (like a piece of plastic wrap) on top of the ground. If the sheet touches the ground at a peak, you check the slope of the sheet. If the sheet is "viscous" (sticky), it won't slide off easily; it tells you where the peak is, even if the ground underneath is jagged.
• Flux-Limited: This is the rule at the intersection. It says, "No matter how fast cars are coming from the left or right, the intersection can only let a certain number of cars through per second."
• t-Measurable: This acknowledges that the "traffic cop" at the intersection and the road conditions are erratic. They aren't smooth curves; they are jagged lines.

How the Author Solved It:
Instead of trying to solve the messy problem directly, Briani uses a clever trick:

1. The Approximation Strategy: She imagines replacing the chaotic, jagged traffic rules with a series of smooth, perfect rules that get closer and closer to the real, messy ones.
2. The Bridge: She proves that if you solve the problem for the smooth, perfect rules, and then slowly make those rules more and more chaotic (approaching the real mess), the solution you get doesn't break. It stays stable.
3. The Optimal Control Connection: She also connects this to a game of "Optimal Control." Imagine a driver trying to get from point A to point B in the shortest time, but the road conditions are random. She proves that the "best possible time" the driver can achieve is exactly the solution to her new equation.

Why This Matters:

• Realism: Real-world systems (traffic, stock markets, heat flow in materials) are rarely smooth. They have sudden jumps, noise, and erratic behavior. This paper gives mathematicians a robust tool to model these real-world messes.
• Traffic Flow: The specific setup (two roads meeting) is a model for traffic jams at intersections. If the traffic lights are broken or erratic, this math helps predict how the jam will form and clear.
• Future Proofing: The author shows that this method isn't just for two roads. It can be expanded to complex city grids (networks) with many intersections, and even to situations where the "rules" depend on the number of cars currently on the road.

In a Nutshell:
This paper is like inventing a new kind of GPS navigation system. Old GPS systems assumed roads were always smooth and traffic lights were perfect. This new system works even when the roads are full of potholes, the traffic lights are flickering randomly, and the traffic cop at the intersection is having a bad day. It proves that even in chaos, there is a predictable pattern if you know how to look for it.

Technical Summary

Here is a detailed technical summary of the paper "Hamilton–Jacobi Equations with Time Measurable Hamiltonians Posed on a 1-Dimensional Junction" by Ariela Briani.

1. Problem Statement

The paper investigates Evolutionary Hamilton–Jacobi (HJ) equations posed on a simple network consisting of two edges (the positive and negative real lines) connected at a single junction ($x=0$). The primary novelty lies in addressing two distinct types of discontinuities simultaneously:

1. Time Discontinuity: The Hamiltonians $H_i(t, x, p)$ and the flux limiter $A(t)$ at the junction are assumed to be measurable in time ($t$) rather than continuous. Specifically, $H_i(\cdot, x, p) \in L^1_{loc}$ and $A \in L^\infty$.
2. Spatial Discontinuity: The problem is defined on a network where the dynamics and costs differ on the two edges ($\Omega_1 = (0, \infty)$ and $\Omega_2 = (-\infty, 0)$), creating a spatial discontinuity at the junction.

The specific model problem considered is:
$$\begin{cases} u_t + H_1(t, x, u_x) = 0 & \text{in } (0, \infty) \times (0, T) \\ u_t + H_2(t, x, u_x) = 0 & \text{in } (-\infty, 0) \times (0, T) \\ u_t(0, t) + \max\{A(t), H_1^-(t, 0, u_x(0^+, t)), H_2^+(t, 0, u_x(0^-, t))\} = 0 & \text{in } (0, T) \\ u(x, 0) = u_0(x) & \text{in } \mathbb{R} \end{cases}$$
where $H_i^\pm$ denote the non-decreasing and non-increasing parts of the convex Hamiltonians.

2. Methodology

The author employs a combination of viscosity solution theory adapted for measurable data and optimal control theory.

A. Definition of Solutions (FL-tm)

The paper introduces the concept of Flux-Limited Time-Measurable (FL-tm) solutions. This definition generalizes two existing frameworks:

• Ishii's definition (1985): For HJ equations with $t$-measurable Hamiltonians in $\mathbb{R}^N$.
• Flux-Limited solutions (Imbert & Monneau, 2010s): For HJ equations on networks with $t$-continuous data.

Key Technical Innovation:
To handle the time measurability, the test functions used in the viscosity definition are modified. A test function $\phi$ is no longer just a smooth function but is decomposed as:
$$\phi(x, t) = \int_0^t b(s) ds + \psi(x, t)$$
where $b \in L^1(0, T)$ accounts for the time discontinuity, and $\psi$ is a piecewise $C^1$ function (continuous across the junction).
The viscosity inequalities at a local extremum $(x_0, t_0)$ are not required to hold directly for the Hamiltonian $H$, but for a continuous function $G$ that "bounds" the Hamiltonian from below (for subsolutions) or above (for supersolutions) in an $L^1$-sense relative to the test function's time derivative.

B. Comparison Principle (Uniqueness)

The proof of the comparison principle (Theorem 3.1) relies on an approximation technique rather than direct doubling of variables, which is difficult with $L^1$ time dependence.

1. Approximation: The measurable Hamiltonians $H_i$ and flux limiter $A$ are approximated by sequences of continuous functions $H_{i,n}$ and $A_n$ converging in $L^1$.
2. Correction Terms: Sub- and supersolutions $u$ and $v$ are perturbed by integrals of the approximation error ($k_n(t) = \sup |H - H_n|$) to construct classical viscosity sub/supersolutions for the approximate continuous problems.
3. Limit Passage: Since comparison holds for the continuous approximations (by existing literature), the limit as $n \to \infty$ yields the comparison result for the measurable case.

C. Existence via Optimal Control

The existence of a solution is established by constructing an optimal control problem.

• Dynamics: Controlled trajectories can move along the edges or stay at the junction (using a control $A_0$).
• Cost: The running cost involves the Lagrangian dual to the Hamiltonians on the edges and a specific cost $-A(t)$ when the trajectory is at the junction.
• Value Function: The value function of this control problem is proven to be the unique FL-tm solution. The proof utilizes the Dynamic Programming Principle (DPP) and verifies the viscosity inequalities by analyzing the behavior of optimal trajectories near the junction.

3. Key Contributions

1. Generalization of Viscosity Theory: The paper successfully extends the theory of Flux-Limited solutions to the case where time dependence is merely measurable ($L^1$), a significant step beyond the standard $t$-continuous assumption.
2. Unified Framework: It provides a rigorous definition (Definition 2.3) that unifies Ishii's measurable theory with network-based flux-limited theories.
3. Existence and Uniqueness:
   • Uniqueness: Proven via the comparison principle for convex Hamiltonians under specific regularity and approximability conditions.
   • Existence: Proven by showing the value function of a specific optimal control problem satisfies the FL-tm definition.
4. Handling Junctions with Measurable Flux: The paper resolves the technical difficulty of defining the junction condition when the flux limiter $A(t)$ is discontinuous in time, ensuring the "max" condition at the junction is well-posed in the viscosity sense.

4. Main Results

• Theorem 3.1 (Comparison Principle): Under assumptions of convexity, coercivity, and a specific "approximability" condition (AP), if $u$ is a subsolution and $v$ is a supersolution with $u(x,0) \leq v(x,0)$, then $u \leq v$ everywhere. This guarantees uniqueness.
• Theorem 4.5 (Existence): The value function of the constructed optimal control problem is a (FL-tm) solution to the HJ equation.
• Regularity: The solution is shown to be bounded and Lipschitz continuous in space (uniformly in time).

5. Significance and Generalizations

• Traffic Flow Modeling: The work is motivated by traffic flow models (referencing Cardaliaguet & Souganidis) where flux limiters (representing traffic light changes or bottleneck capacities) may change abruptly or be modeled as measurable functions rather than smooth ones.
• Robustness of the Method: The approximation-based proof strategy is highlighted as robust. The author discusses in Section 5 how this approach can be extended to:
  • Non-convex Hamiltonians: By assuming quasi-convexity.
  • Dependence on $u$: By adding monotonicity conditions.
  • Complex Networks: The definition naturally extends to networks with multiple junctions, provided uniformity assumptions are maintained.

Conclusion

This paper fills a critical gap in the theory of Hamilton–Jacobi equations on networks by rigorously handling time-measurable discontinuities. By adapting the viscosity solution framework to include $L^1$ time dependencies and proving existence/uniqueness via optimal control, it provides a solid mathematical foundation for modeling dynamic systems (like traffic or fluid flow) on networks where control parameters or external forces change abruptly in time.