Existence results for some nonconvex optimization problems governed by nonlinear processes

Publication date 14 February 2000

nonconvex problems optimal control Young measures

Existence theories for optimal control problems involving ordinary differential equations (49J15) Existence theories for optimal control problems involving partial differential equations (49J20) Methods involving semicontinuity and convergence; relaxation (49J45)

Review text

The author considers optimal control problems of the form: \[ \min J(y,u)=\int_0^T (g(t,y(t))+h(t,u(t)))dt, \] when the pair \((y,u)\), the state and control variables, satisfies the differential equation \(y'=G(t,y(t))+H(t,u(t))\), \(y(0)=0\), and the control constraint \(u(t)\in S(t)\). No convexity assumption is assumed with respect to \(u\). An existence theorem is proved when \(g(t,\cdot)\) is uniformly concave with respect to \(y\), and when the nonlinearity \(G(t,\cdot)\) in the state equation is controlled by the coefficient of uniform concavity of \(g(t,\cdot)\). The idea of the proof is the following. Thanks to Young measures, existence for the relaxed problem is obtained by standard arguments. Next it is proved that the mapping \(\nu\mapsto J(y_\nu,\nu)\), which associates the cost corresponding to a Young measure \(\nu\), is concave. As in \textit{E. J. Balder} [SIAM J. Control Optimization 32, No. 3, 890-916 (1994; Zbl 0813.49004)], the existence follows from the Bauer extremum principle. The case of unbounded controls is also treated with \(L^p\)-Young measures. Other applications to the case of elliptic equations is also studied. These results are interesting generalizations of previous known results in the case when \(G(t,\cdot)\) is linear. For the entire collection see [Zbl 0885.00052].

Recommendations

Cited in

(7)

This page was built for publication: Existence results for some nonconvex optimization problems governed by nonlinear processes

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4249804)