Abstract
The goal of this paper is to derive a traffic flow macroscopic model from a microscopic model with a transition function. At the microscopic scale, we consider a first order model of the form “follow the leader” i.e. the velocity of each vehicle depends on the distance to the vehicle in front of it. We consider two different velocities and a transition zone. The transition zone represents a local perturbation operated by a Lipschitz function. After rescaling, we prove that the “cumulative distribution function” of the vehicles converges towards the solution of a macroscopic homogenized Hamilton-Jacobi equation with a flux limiting condition at junction which can be seen as a LWR model.
You have full access to this open access chapter, Download conference paper PDF
We’re sorry, something doesn't seem to be working properly.
Please try refreshing the page. If that doesn't work, please contact support so we can address the problem.
Keywords
- Macroscopic Traffic Flow Models
- Hamilton-Jacobi Equation
- Local Perturbations
- Microscopic Model
- Flux Limit
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.
1 Introduction
The goal of this paper is to present a rigorous derivation of a traffic flow macroscopic model by homogenization of a follow-the-leader model, see [8, 10]. The idea is to rescale the microscopic model, which describes the dynamics of each vehicle individually, in order to get a macroscopic model which describes the dynamics of density of vehicles. Several studies have been done about the connection between microscopic and macroscopic traffic flow model. This type of connection is important since it allows us to deduce macroscopic models rigorously and without using strong assumptions. We refer for example to [1,2,3] where the authors rescaled the empirical measure and obtained a scalar conservation law (LWR (Lighthill-Whitham-Richards) model). More recently, another kind of macroscopic models appears. These models rely on the Moskowitz function and make appear an Hamilton-Jacobi equation. This is the setting of our work which is a generalization of [6]. Indeed, authors in [6] considered a single road and one velocity throughout this road with a local perturbation at the origin while we consider two different velocities and a transition zone which can be seen as a local perturbation thats slows down the vehicles. At the macroscopic scale, we get an Hamilton-Jacobi equation with a junction condition at zero and an effective flux limiter. In order to have our homogenization result, we will construct the correctors. The main new technical difficulties comes from the construction of correctors and in particular the gradient estimates are more complicated from that in [6] because the gradient on the left and on the right may differ.
2 The Microscopic Model
In this paper, we consider a “follow the leader” model of the following form
where \(U_{j}\) denotes the position of the j-th vehicle and \(\dot{U}_{j}\) its velocity. The function \(\varphi \) simulates the presence of a local perturbation around the origin which allows us to pass from the optimal velocity function \(V_1\) (on the left of the origin) to \(V_2\) (on the right). We make the following assumptions on \(V_{1}\), \(V_{2}\) and \(\varphi \).
Assumption (A).
-
(A1) \(V_{1},V_{2}:\mathbb {R}\rightarrow \mathbb {R}^+\) are Lipschitz continuous, non-negative and non-decreasing.
-
(A2) For \(i=1,2\), there exists a \(h^{i}_0\in (0,+\infty )\) such that
$$ V_{i}(h)=0 \text { for all } h\le h^i_0. $$ -
(A3) For \(i=1,2\), there exists a \(h^{i}_{max}\in (0,+\infty )\) such that
$$ V_{i}(h)=V_{imax} \text { for all }h\ge h^{i}_{max}. $$ -
(A4) For \(i=1,2\), there exists a real \(p^i_0\in [-1/h^i_0,0)\) such that the function \(p\mapsto pV_{i}(-1/p)\) is decreasing on \([-1/h_0^i,p^i_0)\) and increasing on \([p^i_0,0)\).
-
(A5) The function \(\varphi :\mathbb {R}\rightarrow [0,1]\) is Lipschitz continuous and
$$\begin{aligned} \varphi (x)= {\left\{ \begin{array}{ll} 1 &{} \text {if} \, x\le -r \\ 0 &{} \text {if} \, x>r. \end{array}\right. } \end{aligned}$$
3 The Homogenization Result
We introduce the “cumulative distribution function” of the vehicles:
and we make the following rescaling
\(\rho ^{\varepsilon }\) is a discontinuous solution of the following equation: for \((t,x)\in (0,+\infty )\times \mathbb {R}\),
where the non-local operators \(M_{i}^\varepsilon \) and \(M_{2}^\varepsilon \) are defined by
with
We also assume that the initial condition satisfies the following assumption.
(A0) (Gradient Bound). Let \(k_{0}=\max \left( k_{0}^{1},k_{0}^{2}\right) \) with \(k^{i}_{0}=1/h^{i}_{0}\). The function \(u_{0}\) is Lipschitz continuous and satisfies
We have the following theorem (see [6]).
Theorem 1
Assume (A0) and (A). Then, there exists a unique viscosity solution \(u^{\varepsilon }\) of (3.1). Moreover, the function \(u^{\varepsilon }\) is continuous and there exists a constant K such that
We will introduce now the macroscopic model which is a Hamilton-Jacobi equation on a junction. The Hamiltonians \(\overline{H}_{1}\) and \(\overline{H}_{2}\) are called effective Hamiltonians (see Proposition 2.9 in [6]) and are defined as follows: for \(i=1,2\)
with
Now we can define the limit problem. We refer to [9] for more details about existence and uniqueness of solution for this type of equation.
where \(\overline{A}\) has to be determined and \(F_{\overline{A}}\) is defined by
\(\overline{H}_{1}^{+}\) and \(\overline{H}_{2}^{-}\) represent respectively the increasing and the decreasing part of \(\overline{H}_{1}\) and \(\overline{H}_{2}\). The following theorem is our main result in this paper.
Theorem 2
There exists \(\overline{A} \in \left[ H^{1}_0,0\right] \) such that the function \(u^\varepsilon \) defined by Theorem 1 converge locally uniformly towards the unique solution \(u^0\) of (3.6).
Remark 1
Formally, if we derive (3.6), we will obtain a scalar conservation law with discontinuous flux whose literature is very rich, see for example [4]. However, the passage from microscopic to macroscopic models are more difficult in this setting and in particular on networks. On the contrary, the approach proposed in this paper can be extended to models posed on networks (see [5]).
4 Correctors for the Junction
The key ingredient to prove the convergence result is to construct correctors for the junction. Given \(\overline{A} \in \mathbb {R}\), we introduce two real numbers \(\overline{p}_{1}, \overline{p}_{2}\in \mathbb {R}\), such that
If \(\overline{A}\le H_{0}\), we then define \(\overline{p}_{1}, \overline{p}_{2}\in \mathbb {R}\) as the two real numbers satisfying
Due to the form of \(\overline{H}_{1}\) and \(\overline{H}_{2}\) this two real numbers exist and are unique. We consider now the following problem: find \(\lambda \in \mathbb {R}\) such that there exists a solution w of the following global-in-time Hamilton-Jacobi equation
with
Theorem 3 (Existence of a global corrector for the junction)
Assume (A).
-
(i)
(General properties) There exists a constant \(\bar{A}\in [H^{1}_0,0]\) such that there exists a solution w of (4.3) with \(\lambda =\overline{A}\) and such that there exists a constant \(C>0\) and a globally Lipschitz continuous function m such that for all \(x\in \mathbb {R}\),
$$\begin{aligned} |w(x)- m(x)| \le C. \end{aligned}$$(4.5) -
(ii)
(Bound from below at infinity) If \(\bar{A}>H^{1}_0\), then there exists \(\gamma _0\) such that for every \(\gamma \in (0,\gamma _0)\), we have
$$\begin{aligned} \left\{ \begin{array}{l l} w(x - h) - w(x) \ge ( -\overline{p}_{1} - \gamma )h - C &{} \text{ for } x\le -r \text{ and } h\ge 0,\\ w(x+h) - w(x) \ge (\overline{p}_{2} - \gamma )h - C &{} \text{ for } x\ge r \text{ and } h\ge 0.\\ \end{array} \right. \end{aligned}$$(4.6) -
(iii)
(Rescaling w) For \(\varepsilon >0\), we set
$$\begin{aligned} w^\varepsilon (x)= \varepsilon w\left( \dfrac{x}{\varepsilon }\right) , \end{aligned}$$
then (along a subsequence \(\varepsilon _n \rightarrow 0\)) we have that \(w^\varepsilon \) converges locally uniformly towards a function \(W=W(x)\) which satisfies
In particular, we have (with \(W(0)=0\))
5 Proof of Theorem 3
This section contains the proof of Theorem 3. To do this, we will construct correctors on truncated domains and then pass to the limit as the size of the domain goes to infinity. For \(l\in (r,+\infty )\), \(r<<l\) and \(r\le R<<l\), we want to find \(\lambda _{l,R}\), such that there exists a solution \(w^{l,R}\) of
with
and \(\psi _R,\varPhi _{R} \in C^\infty \), \(\psi _R,\varPhi _{R}: \mathbb {R} \rightarrow [0,1]\), with
Proposition 1 (Existence of correctors on truncated domains)
There exists a unique \(\lambda _{l,R} \in \mathbb {R}\) such that there exists a solutions \(w^{l,R}\) of (5.1). Moreover, there exists a constant C (depending only on \(k_0\)), and a Lipschitz continuous function \(m^{l,R}\), such that
Proof
We only give the main steps of the proof. Classically, we will consider the approximated problem depending on the parameter \(\delta \) and then take \(\delta \) to 0.
-
Step 1: construction of barriers. Using Perron’s method and 0 and \(\delta ^{-1}|H^{1}_{0}|\) as barriers, we deduce that there exists a continuous viscosity solution \(v^\delta \) of (5.6) which satisfies
$$\begin{aligned} 0 \le v^\delta \le \dfrac{|H_{0}^{1}|}{\delta }. \end{aligned}$$(5.7) -
Step 2: control of the space oscillations of \(v^\delta \). The function \(v^\delta \) satisfies for all \(x,y\in [-l,l]\), \(x\ge y\),
$$\begin{aligned} -k_{0}(x-y) -1 \le v^\delta (x) - v^\delta (y) \le 0, \end{aligned}$$with \(k_{0}=\max (k_{0}^{1},k_{0}^{2})\) (see [6, Lemma 6.5]).
-
Step 3: construction of a Lipschitz estimate. As in [6, Lemma 6.6] we can construct a Lipschitz continuous function \(m^\delta \), such that there exists a constant C, (independent of l, R and \(\delta \)) such that
$$\begin{aligned} \left\{ \begin{array}{ll} |m^\delta (x) - m^\delta (y)| \le C|x-y| &{} \text{ for } \text{ all } x,y\in [-l,l],\\ |v^\delta (x) - m^\delta (x) |\le C &{} \text{ for } \text{ all } x \in [-l,l]. \end{array} \right. \end{aligned}$$(5.8) -
Step 4: passing to the limit as \(\delta \) goes to 0. Classicly, taking \(\delta \) to zero, we get \(\lambda _{l.R}, w^{l,R}\) and \(m^{l,R}\) satisfiying (5.5). The uniqueness of \(\lambda _{l,R}\) is classical so we skip it. This ends the proof of Proposition 1. \(\square \)
Proposition 2
The following limits exist (up to a subsequence)
Moreover, we have
Proposition 3 (Control of the slopes on a truncated domain)
Assume that l and R are big enough. Let \(w^{l,R}\) be the solution of (5.1) given by Proposition 1. We also assume that up to a sub-sequence \(\overline{A}=\lim \limits _{R\rightarrow +\infty } \lim \limits _{l\rightarrow +\infty }\lambda _{l,R}>H^{1}_0\). Then there exits a \(\gamma _0>0\) such that for all \(\gamma \in (0,\gamma _0)\), there exists a constant C (independent of l and R) such that for all \(x\le -r\) and \(h\ge 0\)
Similarly, for all \(x\ge r\) and \(h\ge 0\),
Proof
We only prove (5.9) since the proof for (5.10) is similar. For \(\sigma >0\) small enough, we denote by \(p_-^\sigma \) the real number such that
Let us now consider the function \(w^-= p_-^\sigma x\) that satisfies
We also have
For all \(x\in (-l,-r)\), using that \(\varphi (x)=1\) and \(\psi _{R}(x)=1\), we deduce that \(w^{-}\) satisfies
Using the comparaison principle, we deduce that for all \(h\ge 0\), for all \(x\in (-l,-r)\), we have that
Finally, for \(\gamma _{0}\) and \(\sigma \) small enough, we can set \(p^{\sigma }_{-}=\overline{p}_{1}+\gamma \). \(\square \)
Proof of Theorem
3. The proof is performed in two steps.
Step 1: proof of (i) and (ii). The goal is to pass to the limit as \(l\rightarrow +\infty \) and then as \(R\rightarrow +\infty \). There exists \(l_n \rightarrow +\infty \), such that
the convergence being locally uniform. We also define
Thanks to (5.5), we know that \(\overline{w}^R\) and \(\underline{w}^R\) are finite and satisfy
By stability of viscosity solutions, \(\overline{w}^R-2C\) and \(\underline{w}^R\) are respectively a sub and a super-solution of
Therefore, using Perron’s method, we can construct a solution \(w^R\) of (5.11) with \(m^R, \overline{A}^R\) and \(w^R\) satisfying
Using Proposition 3, if \(\overline{A}>H_0\), we know that there exists \(\gamma _0\) and \(C>0\), such that for all \(\gamma \in (0,\gamma _0)\),
Passing to the limit as \(R\rightarrow +\infty \) and proceeding as above, the proof is complete.
Step 2: proof of (iii). Using (4.6), we have that
Therefore, we can find a sequence \(\varepsilon _n \rightarrow 0\), such that
with \(W(0)=0\). Like in [7](Appendix A.1), we have that
For all \(\gamma \in (0,\gamma _0)\), we have that if \(\overline{A}>H^{1}_0\) and \(x>0\),
where we have used (4.6). Therefore we get
Similarly, we get \( W_x = \overline{p}_1\) for \(x<0\). This ends the proof of Theorem 3. \(\square \)
6 Proof of Convergence
In this section, we will prove our homogenization result. Classicly, the proof relies on the existence of correctors. We will just prove the convergence result at the junction point since at any other point, the proof is classical using that \(v=0\) is a corrector, see [6].
Proof of Theorem
2. We introduce
Let us prove that \(\overline{u}\) is a sub-solution of (3.6) at the point 0, (the proof for \(\underline{u}\) is similar and we skip it). The definition of viscosity solution for Hamilton-Jacobi equation is presented in Sect. 2 in [9]. We argue by contradiction and assume that there exist a test function \(\varPsi \in \mathcal {C}^1(J_\infty )\) such that
According to [9], we may assume that the test function has the following form
The last line in condition (6.2) becomes
Let us consider w the solution of (4.3) provided by Theorem 3, and let us denote
We claim that \(\varPsi ^{\varepsilon }\) is a viscosity solution on \(Q_{\bar{r},\bar{r}}(\bar{t},0)\) of the following problem,
Indeed, let h be a test function touching \(\varphi ^\varepsilon \) from below at \((t_1,x_1)\in \mathcal {Q}_{\bar{r},\bar{r}}(\bar{t},0)\), so we have that the function \(\chi (y)=\dfrac{1}{\varepsilon }\left( h(t_{1},\varepsilon y)-g(t_{1})\right) \) touches w from below at \(\dfrac{x_{1}}{\varepsilon }\) which implies that
Using (6.4) and the fact that \(h_{t}(t_{1},x_{1})=g^{\prime }(t_{1})\) and computing (6.6), we get the desired result.
Getting the Contradiction. We have that for \(\varepsilon \) small enough
Using the fact that \(w^\varepsilon \rightarrow W\), and using (4.8), we have for \(\varepsilon \) small enough
Combining this with (6.5), we get that
By the comparison principle on bounded subsets the previous inequality holds in \(\mathcal {Q}_{\bar{r},\bar{r}}(\bar{t},0)\). Passing to the limit as \(\varepsilon \rightarrow 0\) and evaluating the inequality at \((\bar{t},0)\), we obtain the following contradiction
\(\square \)
References
Aw, A., Klar, A., Rascle, M., Materne, T.: Derivation of continuum traffic flow models from microscopic follow-the-leader models. SIAM J. Appl. Math. 63, 259–278 (2002)
Colombo, R.M., Rossi, E.: Rigorous estimates on balance laws in bounded domains. Acta Math. Sci. 35, 906–944 (2015)
Di Francesco, M., Rosini, M.D.: Rigorous derivation of the lighthill-whitham-richards model from the follow-the-leader model as many particle limit. arXiv preprint arXiv:1404.7062 (2014)
Diehl, S.: On scalar conservation laws with point source and discontinuous flux function. SIAM J. Math. Anal. 26, 1425–1451 (1995)
Forcadel, N., Salazar, W.: Homogenization of a discrete model for a bifurcation and application to traffic flow (2016). hal-01332787
Forcadel, N., Salazar, W.: A junction condition by specified homogenization of a discrete model with a local perturbation and application to traffic flow, \(<\)hal-01097085\(>\) (2015)
Galise, G., Imbert, C., Monneau, R.A: Junction condition by specified homogenization. arXiv preprint arXiv:1406.5283 (2014)
Gazis, D.C., Herman, R., Rothery, R.: Nonlinear follow-the-leader models of traffic flow. Oper. Res. 9, 545–567 (1961)
Imbert, C., Monneau, R.: Flux-limited solutions for quasi-convex Hamilton-Jacobi equations on networks, \(<\)hal-00832545\(>\) (2014)
Prigogine, I. Herman, R.: Kinetic theory of vehicular traffic, p. 100 (1971)
Acknowledgments
This work is co-financed by the European Union with the European regional development fund (ERDF, HN0002137) and by the Normandie Regional Council via the M2NUM project.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2016 IFIP International Federation for Information Processing
About this paper
Cite this paper
Forcadel, N., Zaydan, M. (2016). Derivation of a Macroscopic LWR Model from a Microscopic follow-the-leader Model by Homogenization. In: Bociu, L., Désidéri, JA., Habbal, A. (eds) System Modeling and Optimization. CSMO 2015. IFIP Advances in Information and Communication Technology, vol 494. Springer, Cham. https://doi.org/10.1007/978-3-319-55795-3_25
Download citation
DOI: https://doi.org/10.1007/978-3-319-55795-3_25
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-55794-6
Online ISBN: 978-3-319-55795-3
eBook Packages: Computer ScienceComputer Science (R0)