Modeling of crowds in regions with moving obstacles (Q1983295)

The authors propose a model accounting for crowds moving in domains with obstacles and written as: \(\partial _{t}\rho _{t}+\nabla \cdot (v_{t}\rho _{t})=0\), \(v_{t}(x)-\mathcal{V}(\rho _{t})(x)\in -N_{C(t)}(x)\), \(spt(\rho _{t})\subset C(t)\), where \(v_{t}\) is the velocity, \(\rho _{t}\) the density, \( C(t)=\mathbb{R}^{d}\setminus O(t)\) is the complement in \(\mathbb{R}^{d}\) of the obstacle \(O(t)\) (\(C(t)\) is the viability region), \(\mathcal{V}\) maps measures to vector fields and \(N_{A}(x)\) is the proximal normal cone to \( A\subset \mathbb{R}^{d}\) at \(x\). Defining \(\mathcal{P}_{2}(\mathbb{R}^{d})\) as the space of probability measures with finite second order moments equipped with the Wasserstein distance \(W_{2}\), an absolutely continuous curve \(\rho :[0,T]\rightarrow \mathcal{P}_{2}(\mathbb{R}^{d})\) is a solution to the above problem if there exists a Borel vector field \((t,x)\rightarrow v_{t}(x)\) such that \((\rho ,v)\) satisfies the first above equation in the sense of distributions, the normal cone condition expressed in the second above equation holds for a.e. \(t\in \lbrack 0,T]\) and \(\rho _{t}\)-a.e. \(x\in \mathbb{R}^{d}\), and the support \(spt(\rho _{t})\) of \(\rho _{t}\) is included in \(C(t)\) for all \(t\in \lbrack 0,T]\). The main result of the paper proves that for any \(\vartheta \in \mathcal{P}_{2}(\mathbb{R}^{d})\) such that \( spt\vartheta \subset C(0)\), there exists a unique solution \(\rho :[0,T]\rightarrow \mathcal{P}_{2}(\mathbb{R}^{d})\) to the above problem with \(\rho (0)=\vartheta \). The authors here assume Lipschitz properties of \( \mathcal{V}\) and \(C\) and they prove further properties of this solution. For the proof, the authors recall properties of spaces of measures and of the Wasserstein distance. For the existence part of the proof, the authors build processes through a continuous or a piecewise continuous approximation and they establish properties of these processes. They also use properties of the Benamou-Brenier functional that they recall in an Appendix. For the uniqueness part, they prove a continuous dependence property and they apply Gronwall's Lemma. In the last parts of their paper, the authors prove that it is possible to place an obstacle so that the number of agents staying in a ''dangerous'' zone is minimal. They finally present results of numerical simulations, considering two examples: an attraction/repulsion model and a congestion model.