Measure-valued solutions to nonlocal transport equations on networks (Q1708133)

The authors consider a network \(\Gamma =(\mathcal{V},\mathcal{E})\) defined as a finite collection of vertices \(\{x_{i}\}_{i\in I}\) and a finite collection of continuous non-self intersecting arcs \(\{e_{j}\}_{j\in J}\) whose endpoints belong to \(\mathcal{V}\). \(\mathcal{S}\) is the set of sources that is the vertices where the particles enter in the network. The main result of the paper proves the existence of a unique solution to the nonlinear transport problem \(\partial _{t}m+\partial _{x}(v[m_{t}]m)=0\) posed in \(\Gamma \times [ 0,T]\) with the initial condition \( m_{t=0}=m_{0}\) and the boundary conditions \(m_{x=x_{i}}=\sigma _{0}^{i}\) for every \(x_{i}\in \mathcal{S}\) and \(m_{x=x_{i}}^{j}=\sum_{k\in \mathrm{Inc}(x_{i})}p_{kj}\cdot m_{x=x_{i}}^{k}\) for every \(x_{i}\in \mathcal{V} \setminus \mathcal{S}\) and \(e_{j}\in \mathrm{Out}(x_{i})\). Here \(\mathrm{Inc}(x_{i})\) and \(\mathrm{Out}(x_{i})\) respectively denote the set of outgoing or incoming arcs in \( x_{i}\), \(\sigma _{0}\) is a boundary measure \(\sigma _{0}=\sum_{x_{i}\in \mathcal{S}}\sigma _{0}^{i}\), where \(\sigma ^{i}\in \mathcal{M} ^{+}(\{x_{i}\}\times [ 0,T])\), \(m_{0}\) is the initial mass distribution \(m_{0}=\sum_{j\in J}m_{0}^{j}\) with \(supp\) \(m_{0}^{j}\sqsubseteq e_{j}\) and the matrix \(p_{kj}\) is a stochastic one: \(0<p_{kj}\leq 1\), \( \sum_{j=1,J}p_{kj}=1\) and \(p_{kj}=0\) if \(e_{k}\cap e_{j}=\varnothing \) or if \(e_{j}\) comes before \(e_{k}\)\ in the assigned orientation of the network. The function \(v:\mathcal{M}^{+}(\Gamma )\times \Gamma \rightarrow \mathbb{R}\) is nonnegative, has a positive maximum, is Lipschitz continuous with respect to the state variable and is Lipschitz continuous with respect to the measure (with the norm \(\left\| \mu \right\| _{BL}^{\ast }=\sup_{\phi \in BL(\mathcal{T}),\left\| \phi \right\| _{BL}\leq 1}\), \(BL(\mathcal{T })\) being the Banach space of bounded and Lipschitz continuous functions on the topological space \(\mathcal{T}\)). The authors first define the notion of measure-valued solution \(m\in \mathcal{M}^{+}(\Gamma \times [ 0,T])\) if it satisfies \(\left\langle m_{t=T}-m_{0},f\right\rangle -\left\langle \sigma _{0},f\right\rangle =\left\langle m,\partial _{t}f+v(x)\partial _{x}f\right\rangle \) for every \(f\in C_{0}^{1}(\Gamma \times [ 0,T])\) and \(\left\langle m_{x=x_{i}}^{j},f\right\rangle =\sum_{k\in \mathrm{Inc}(x_{i})}\left\langle m_{x=x_{i}}^{k},p_{kj}f\right\rangle \) for every \( x_{i}\in \mathcal{V}\setminus \mathcal{S}\) and \(e_{j}\in \mathrm{Out}(x_{i})\). Under appropriate hypotheses on the initial data \(m_{0}\) and \(\sigma _{0}\), the authors prove the existence of a measure \(m\) to the above nonlinear transport problem. For the proof, the authors introduce the intervals \( I_{n}^{N}=(t_{n}^{N},t_{n+1}^{N})\) with \(t_{n}^{N}=nT/2^{N}\) and the problem \(\partial _{t}m+\partial _{x}(v[m_{t_{n}}]m)=0\) posed in \(\Gamma \times I_{n}^{N}\) with the conditions \(m_{t=t_{n}}=m_{t_{n}}\), \(m_{x_{i}\in \mathcal{S}}=\sigma _{0}\llcorner I_{n}^{N}\), \(m_{x=x_{i}}^{j}=\sum_{k\in \mathrm{Inc}(x_{i})}p_{kj}\cdot m_{x=x_{i}}^{k}\) for every \(x_{i}\in \mathcal{V} \setminus \mathcal{S}\) and \(e_{j}\in \mathrm{Out}(x_{i})\), where \(\sigma _{0}\llcorner I_{n}^{N}\) means the restriction of \(\sigma _{0}\) to \(I_{n}^{N} \). The authors observe that \(v[m_{t_{n}}]\) is linear on \(\Gamma \times I_{n}^{N}\) and they quote from their paper [Netw. Heterog. Media 12, No. 2, 191--215 (2017; Zbl 1364.35400)] the existence of a solution \(m^{N,n}\in \mathcal{M}^{+}(\Gamma \times I_{n}^{N})\) to this linear approximate problem. They define \(m_{t}^{N}=m_{t}^{N,n}\) on \(I_{n}^{N}\) and they prove that \(m_{t}^{N}\) is bounded in \((\mathcal{M}^{+}(\Gamma ),\left\| \cdot \right\| _{BL}^{\ast })\) uniformly with respect to \(N\) and that \( \left\| m_{t}^{N}-m_{s}^{N}\right\| _{BL}^{\ast }\leq \sigma _{0}([s,t])+C\left| t-s\right| \), for \(0\leq s<t\leq T\). They finally prove that the measure \(m^{N}\) converges to a measure \(m\) which is a solution to the above nonlinear transport problem. The second main result of the paper proves that this solution depends in a continuous way on the initial data \(m_{0}\) and \(\sigma _{0}\). The paper ends with an application to a traffic problem on networks.