An exact dual method of solution of a linear terminal control problem (Q797132): Difference between revisions

The paper deals with the following linear optimal control problem: \[ (1)\quad J(u)=c'x(t_ 1)\to \max;\quad \dot x=Ax+bu,\quad x(0)=x_ 0,\quad Hx(t_ 1)=g,\quad d_*\leq u(t)\leq d^*,\quad t\in T=[0,t_ 1]. \] Problem (1) is reduced to a linear programming problem in a Banach space, then its dual problem is considered. The authors present an algorithm of transformation of the initial information of the dual problem for obtaining an optimal control \(u^ 0(t)\), \(t\in T\), or an \(\epsilon\)-optimal control \(u^{\epsilon}(t)\), \(t\in T\) for problem (1). The method presented is exact, that is it involves not estimates but the solutions of the dual problem. This method is based on linear programming methods described by \textit{R. F. Gabasov} and \textit{F. M. Kirillova} [''Linear programming methods'', I-III (Russian) (1977/78/80; Zbl 0468.90035, Zbl 0468.90036 and Zbl 0484.90067)].