Attainability property for a probabilistic target in Wasserstein spaces (Q2229368)

In this paper, the authors study the attainability problem for multi-agent systems in \(\mathbb{R}^d\), i.e., where the number of agents is so large that only a macro level description is feasible, with \textit{nonlocal} interaction between the agents. The system is assumed to be isolated, i.e., the total number of agents is constant in time, and the evolution is thus governed by a continuity-type equation \[ \partial_t \mu_t +\nabla\cdot(v_t\mu_t)=0, \] where \(\mu_t \) describes the distribution of agents, and \(v_t\) is some velocity field induced during the evolution. That is, for a given set \(A\subseteq\mathbb{R}^d\), the ratio \(\mu_t(A)/\mu_t(\mathbb{R}^d)\) is the fraction of agents inside \(A\). Since the system is isolated, it is not restrictive to assume \(\mu_t(\mathbb{R}^d)\equiv 1\), and the Wasserstein space is thus the most natural space to work in. A ``trajectory'' is a measure valued mapping \(\nu\) satisfying the continuity equation, and such trajectory is ``admissible'' if, roughly speaking, it satisfies the continuity equation, and at every time \(t\) the measure \(\nu_t\) is absolutely continuous with respect to \(\mu_t\). The authors study the attainability problem, i.e., given a target configuration, is it possible to reach such it through an admissible trajectory? And if so, what is the minimum time to reach it? The authors provide positive results to both questions, with minimal additional hypotheses. Then they estimate the time to reach a target configuration, and the regularity of such maps. Finally, they perform a comparison with the attainability results in the classical, finite dimensional setting.