scientific article; zbMATH DE number 1273994

From MaRDI portal
Publication:4237483





The author studies the optimal control problem: \(\dot x=f(x,u,t),\) \(t\in [t_1,t_2]\), \(u(t)\in U(x,t)\) \(\dot \forall t\), \(e_i(p)\leq 0\) \(\forall i\in I_1\), \(e_i(p)=0\) \(\forall i\in I_2\), \(p=(t_1,t_2,x_1,x_2)\), \(x_1=x_1(t_1)\), \(x_2=x_2(t_2)\), \(J=J(u,p)=e_0(p)\to \min\) (symbol \(\dot \forall\) means ``for almost all). The multivalued mapping \(U\) puts every pair \((x,t)\) into correspondence with the nonempty compact set \(U(x,t)\subset R^m\). The set of various neasurable essentially bounded functions \(u(.)\) is a class of admissible functions. The pair \((w,p)\) is said to be an admissible process if the local unique solution \(x(.)\) can be extended to the whole interval \([t_1,t_2]\) and the terminal constraints are satisfied as well. The example of the nonexistence of the optimal process is given. In order to assure the existence of the optimal process the class of generalized controls in a form of a family of Radon probability measure is introduced. The class of generalized controls naturally includes the class of ordinary controls due to the identification of the ordinary control \(u\) with the generalized control \(\delta_{u(t)},\) where \(\delta_v\) is a Dirac measure concentrated at the point \(v\). The main result is the theorem about the existence of the approximated optimal process in the class of generalized controls and its relation to the original optimal control problem. For the entire collection see [Zbl 0910.00033].











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4237483)