Second Order Sufficiency Criteria for a Discrete Optimal Control Problem

From MaRDI portal





In this paper the authors prove a large number of results concerning necessary and/or sufficient second order optimality conditions for a discrete-time optimal control problem which consists in the minimization of the functional \[ {\mathcal F}(x,u):=K(x_0,x_{N+1})+\sum_{k=0}^N g(k,x_{k+1},u_k) \] subject to: \[ x_{k+1}-x_k=f(k,x_{k+1},u_k), \;\varphi (x_0,x_{N+1})=0, \;x_k\in \mathbb{R}^n \] \[ \psi_k(u_k)=0, \;u_k\in \mathbb{R}^m, \;k\in J:=\{0,1,\dots \;N\} \] where the data are smooth mappings whose derivatives satisfy certain rank conditions. Applying the Lagrange multiplier rule, the authors obtain first a ``discrete maximum principle then try to extend to this problem the Legendre and Jacobi theories in the classical calculus of variations to obtain necessary and/or sufficient second order optimality conditions. To this end, the authors introduce the second variation of the functional, the accessory problem, the accessory system (which turns out to be a Hamiltonian linear difference system), the associated matrix Riccati difference equations and corresponding concepts of conjugacy and focal points which are used to obtain several characterizations of the (strict) positivity of the second variation.




Cited in
(32)








This page was built for publication: Second Order Sufficiency Criteria for a Discrete Optimal Control Problem

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3148759)