Limits of Singularly Perturbed Control Problems with Statistical Dynamics of Fast Motions

From MaRDI portal





A singularly perturbed control problem for the system of two equations \[ \begin{cases}\dot x=f(u,x,y)\\ \varepsilon\dot y=g(u,x,y),\quad x(0)=x_0,\quad y(0)=y_0\end{cases}\tag{1a} \] with the controls \(u(t)\in U\) and the Lagrange cost functional \[ J_\varepsilon(u)=\int_0^1 Q(u(t),x_\varepsilon(t),y_\varepsilon(t)) dt\tag{1b} \] (where \((x_\varepsilon,y_\varepsilon)\) is the assumed unique solution to (1a)) is considered. For this problem the limit behavior of the trajectories and the cost (as the small parameter \(\varepsilon\) tends to zero) as well as the construction of a limit control system are studied. The main result consists in convergence theorems for trajectories obtained for different classes of controls (continuous, piecewise continuous and measurable ones) and the theorem on the continuity of the value function of the control problem for (1a), (1b) with respect to \(\varepsilon\) on the families of so-called relatively slow controls. It is shown that if there is an invariant measure \(\nu(u,x)\) for \({dy\over d\tau}=g(u,x,y)\), then the limit control problem is of the form \[ \dot x=\int_{{\mathbb{R}}^m}f(u,x,y)\nu(u,x)(dy),\quad x(0)=x_0\tag{2a} \] and \[ J_0(u)=\int_0^1 Q(u(t),x_0(t),y)\mu_0(t) (dy),\tag{2b} \] where \(\mu(t)=\nu(u(t),x(t))\) and \((x_0(t),\mu_0(t))\) denotes the trajectory of (2a) generated by a control \(u(t)\). The limit problem (2a), (2b) approximates in some sense the problem (1a), (1b). The approach developed in this paper uses some properties of dynamical systems and it is based on the average technique combined with the relaxation method introduced originally by J. Warga and developed later by Z. Artstein.











This page was built for publication: Limits of Singularly Perturbed Control Problems with Statistical Dynamics of Fast Motions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3128445)