No label defined (Q3760414)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 4010368
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | No label defined |
scientific article; zbMATH DE number 4010368 |
Statements
1987
0 references
singular perturbations
0 references
Bellman equation
0 references
ergodic control problems
0 references
0.9048638343811036
0 references
0.8601303100585938
0 references
0.8471471071243286
0 references
[For the entire collection see Zbl 0605.00020.] The following class of control problems is considered where the system is governed by: \[ dx=f(x,y,v)dt+\sqrt{2}dw\quad (slow\quad system),\quad \epsilon dy=g(x,y,v)dt+\sqrt{2\epsilon}db\quad (fast\quad system) \] with b and w independent Wiener processes. There is full information and the objective is to minimize the payoff \[ J^{\epsilon}_{x,y}(v)=E\int^{\tau}_{0}e^{-\beta t} \ell (x^{\epsilon}(t), y^{\epsilon}(t))dt \] where \(\tau\) denotes the first exit time of the process x from the boundary of a domain \({\mathcal O}\) (smooth, bounded). The value \(u^{\epsilon}(x,y)\) is the classical solution of a Bellman equation. Proceeding formally, if \[ L(y,v)=\ell (x,y,v)+pf(x,y,v),\quad G(y,v)=g(x,y,v),\quad H(y,q)=\inf_{v\in {\mathcal V}_{\alpha d}}[L(y,v)+qG(y,v)] \] and if there exists a constant \(\chi\) (with respect to y) and a function \(\phi\) such that (1) \(-\Delta_ y\phi +\chi =H(y,D_ y\phi)\), any solution u of (2) \(-\Delta u+\beta u=\chi (x,D_ xu)\) vanishing on the boundary of the domain \({\mathcal O}\) can be an expected limit of the \(u^{\epsilon}.\) Consequently this procedure depends on the possibility to solve ergodic control problems of type (1). If v(y) is a feedback control and (3) \(dy=G(y,v(y))dt+\sqrt{2}db\), \(y(0)=y\) the ergodicity of y must be required. This property is achieved when G is periodic in y or when instead of (3) a reflected diffusion is considered. In sections 1 or 4 these two cases are studied. In section V a fast system \[ dy=G(v)y dt+\sum^{\bar m}_{m=1}\sigma_ m y db_ m(t),\quad y(0)=y \] is considered, and if \(\rho (t)=| y(t)|\), \(\xi (t)=y(t)/| y(t)|\), the slow system is of the form \[ dx=f(x,\rho,\xi,v)dt+\sqrt{2}dw,\quad x(0)=x. \] The cost function to minimize is \(E\int^{\tau}_{0}e^{-\beta t}\ell (x,\rho,\xi,v)dt\) and this model is studied when \(\rho\to 0\) as \(t\to \infty\), and \(\xi\) is ergodic. Finally in sections 6 and 7, the equation (3) is considered in the whole space (with smooth assumptions to recover the ergodicity).
0 references