Mayer control problem with probabilistic uncertainty on initial positions (Q1688661)

The paper under review considers a finite horizon optimal control problem of Mayer type, set on the space of nonnegative Borel probability measures \(\mathcal{P}_2(\mathbb{R}^d)\) with finite second moment. In this setting the authors aim to model scenarios when the knowledge of the initial state and velocity is subject to some uncertainty. The problem reads as follows: let \(T>0\) be fixed, for a given \(s\in[0,T]\) and \(\mu\in \mathcal{P}_2(\mathbb{R}^d)\) define the value function \(V:[0,T]\times \mathcal{P}_2(\mathbb{R}^d)\to\mathbb{R}\) as \[ V(s,\mu):=\inf\left\{\mathcal{G}(\mu_T):\{\mu_t\}_{t\in[s,T]}\in\mathcal{A}^F_{[s,T]}(\mu)\right\}, \] where \(\mathcal{G}:\mathcal{P}_2(\mathbb{R}^d)\to\mathbb{R}\) represents the final cost, which is a given bounded and Lipschitz continuous map w.r.t. the \(W_2\) Wasserstein metric arising from the theory of optimal transportation; \(\{\mu_t\}_{t\in[s,T]}\in\mathcal{A}^F_{[s,T]}(\mu)\) are admissible curves of probability measures. This latter set is defined as distributional solutions on \((s,T)\times\mathbb{R}^d\) of the continuity equation \[ \partial_t\mu+\nabla\cdot(\mu_t v_t)=0 \] with \(\mu_s=\mu\) and \(v_t(x)\in F(x)\) for a.e. \(t\in[s,T]\) and \(\mu_t-\)a.e. \(x\in\mathbb{R}^d\). The given set-valued function \(F\) is Lipschitz continuous with nonempty convex and compact values. As their main result, the authors show that -- just as in the finite dimensional case -- the value function \(V\) solves a Hamilton-Jacobi equation in a suitable viscosity sense. In addition, this solution is unique. Then, these results are applied to study a zero sum pursuit-evasion game in the same setting. The authors show that this game admits a value, by proving that the upper and the lower values are sub- and supersolutions of the same Hamilton-Jacobi equation. The conclusion is achieved by a comparison principle.