Topics in modern computational methods for optimal control problems

Publication date 21 February 1993

finite identification of active indices gradient projection method quasi-Newton methods

[For the entire collection see Zbl 0746.00042.] This review article focusses on a few aspects of modern developments in the field of numerical solutions for optimal control problems. The second chapter deals with the gradient projection method. This is a very robust but also locally rather slow method. This disadvantage can be overcome by Newton type versions of it. It is shown how for example the convergence rate in the finite dimensional case can be predicted and explained by looking at the underlying infinite dimensional problem. Also, new developments like finite identification of active indices are considered from the point of view of optimal control. The third chapter is devoted to quasi-Newton methods which have been used very successfully in finite dimensional optimization during the past two decades. Their application to optimal control problems can be interpreted much better by the use of recent result in the convergence analysis for infinite dimensional problems. The fourth chapter shows how the description of the dynamics of the system can be used in the design of quasi-Newton methods. This leads to a new method with substantial improvement over existing methods in its convergence features.

Recommendations

Cited in

(6)

This page was built for publication: Topics in modern computational methods for optimal control problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4026995)