Existence of an optimal control for systems of differential equations with pulse action at nonfixed times (Q1703348)

The discussed paper deals with an optimal control problem for a system of differential equations with a pulse action. More precisely, one considers a control system in \(\mathbb{R}^d\), linear in control: \[ \begin{aligned} &\dot x=A(x,t)+B(x,t)u\\ &x(0)=x_0\;, \end{aligned} \] with an additional pulse action whenever \(x(t)\) hits a given hypersurface \(S\subset \mathbb{R}^d\). This pulse is defined by a continuous function \(g:S\rightarrow \mathbb{R}\), i.e. \[ \Delta x\big|_{x\in S}=g(x)\;. \] The considered cost functional reads as \[ J(u)=\int_0^TA^0(x,t)+B^0(u,t)\mathrm{d}t \;. \] The main result of the paper, Theorem 1, states that under mild technical assumptions the above optimal control problem has a solution. The proof, rather than relying on the Maximum Principle, follows a standard scheme of building a solution out of a weakly convergent sequence of approximations, making use of techniques expected in this context (e.g. Gronwall's inequality, Arzela-Ascoli Lemma, Lebesque convergence theorems, various integral estimates, etc.). However, the presence of the pulse action considerably complicates the reasoning. In principle, the convergent sequences are constructed on each interval defined by subsequent \(S\)-return times of the system (i.e. where the puse action is not active) and then carefully glued together. To do this Lemma 1 -- an auxiliary technical result establishing the regularity of \(S\)-return times -- is needed. Despite its technical character, I found the reviewed paper easy to follow, well-written and enjoyable.