root/applications/dual/SIDP/text/ch3.tex @ 891

A�liv pou�it�ynamick� programov� p����ok v ��lohy du�� ��analytick�e�en�bvykle nen�o�n��at. V ka�d��ov�kroku se toti� pot�se dv� obecn�bt��obl�my: 1) v� st� hodnoty a 2) minimalizace vzhledem k $u_t$. Oba probl� obecn�emaj�nalytick�e�en� bez dal��pecifikace � je proto t�p� k aproxima�m metod�

V t� kapitole p��me popis n�lika mo�n��up�proximativn� ��lohy du�� ��P�e� �e �u du�� ��je nalezen��c�trategie $\pi=\mu_{0:N-1}$, kter�y minimalizovala o��nou ztr�
\begin{equation}
\label{ilos}
J_\pi=\E_{y_0,w_{0:N-1}}\left\{g_N(y_N)+\sum_{t=0}^{N-1}g_t(y_t,\mu_t(I_t),w_t)\right\},
\end{equation}
za podm�k
\begin{gather}
\label{the2}
\theta_{t+1}=h_t(\theta_t,I_t,y_{t+1},u_t),\\
\label{poz3}
y_0=h_0(\theta_0,v_0),\qquad y_{t+1}=h_t(\theta_t, I_t,u_t,v_{t+1}), \qquad t=0,\ldots,N-1,
\end{gather}

\section{Certainty equivalecnce control}
\section{Metoda separace}
\section{SIDP}