Abstract:We examine the numerical approximation of time-dependent Hamilton-Jacobi equations on networks, providing a convergence error estimate for the semi-Lagrangian scheme introduced in (Carlini and Siconolfi, 2023), where convergence was proven without an error estimate. We derive a convergence error estimate of order one-half. This is achieved showing the equivalence between two definitions of solutions to this problem proposed in (Imbert and Monneau, 2017) and (Siconolfi, 2022), a result of independent interest, and applying a general convergence result from (Carlini, Festa and Forcadel, 2020).
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to provide a convergence error estimate for the semi - Lagrangian numerical scheme of time - varying Hamilton - Jacobi (HJ) equations on networks. Specifically, the author focuses on how to quantify the accuracy of this numerical method when approximating the solutions of HJ equations defined on networks.
### Background and Problem Description
1. **Research Background**:
- In the past few years, control problems on networks have received extensive attention due to their wide range of application fields (such as data transmission, traffic management, flame propagation, and urban infrastructure planning).
- Solving HJ equations on networks has become very challenging, especially when dealing with the connection conditions at nodes. These nodes are the places where multiple edges intersect and introduce unique complexity.
2. **Main Problems**:
- The network structure brings the need for new mathematical techniques and appropriate numerical methods, especially for dealing with discontinuities at nodes.
- Two main methods have been proposed: one uses special test functions (proposed by Imbert and Monneau), and the other uses standard test functions (proposed by Siconolfi).
- Recent research has proposed a new semi - Lagrangian scheme, which utilizes the viscosity solution framework and allows for larger time steps, but its convergence error estimate has not yet been proven.
### Paper Objectives
The main objective of this paper is to provide a convergence error estimate for the semi - Lagrangian scheme proposed in [7]. To this end, the author takes the following steps:
1. **Prove the Equivalence of Two Viscosity Solution Definitions**:
- The author first proves the equivalence between the two viscosity solution definitions proposed by Imbert and Monneau and Siconolfi respectively. This result is of independent significance in itself because it links two important frameworks.
2. **Apply General Convergence Results**:
- Using the above - mentioned equivalence, the author applies the general convergence results in [6], thereby providing new convergence results and error estimates for the scheme in [7].
3. **Derive Error Estimates**:
- By proving the equivalence of the two viscosity solution definitions, the author is able to apply the general convergence results in [6], thereby obtaining the convergence error estimate of the semi - Lagrangian scheme. The error estimate is \(O(\sqrt{\Delta t})\), where \(\Delta t\) is the time step.
### Conclusions
This paper successfully provides a convergence error estimate for the semi - Lagrangian numerical scheme of time - varying Hamilton - Jacobi equations on networks. By proving the equivalence of the two viscosity solution definitions and applying general convergence results, the author not only fills the gaps in previous research but also provides a theoretical basis for future numerical analysis.
### Formula Summary
- **Error Estimate Formula**:
\[
|v(x, t) - u(x, t)| \leq C T \left( \frac{1}{\sqrt{\Delta t}} \min \left\{ \Delta x, \frac{\Delta x^2}{\Delta t} \right\} + \sqrt{\Delta t} \right)
\]
where \(C\) is a constant independent of \(\Delta\), \(\Delta x\) is the space step, and \(\Delta t\) is the time step.
- **Lagrangian Definition**:
\[
L_\gamma(s, \lambda) = \max_{\mu \in \mathbb{R}} (\lambda \mu - H_\gamma(s, \mu))
\]
- **Modified Hamiltonian**:
\[
H_\gamma(s, \mu) =
\begin{cases}
eH_\gamma(s, \mu), & \text{if } \mu \leq b\mu_\gamma \\
eH_\gamma(s, b\mu_\gamma), & \text{if } \mu > b\mu_\gamma
\end{cases}
\]