Publicou 12 edições por ano
ISSN Imprimir: 0040-2508
ISSN On-line: 1943-6009
Indexed in
The Method Based on the Theory of Atomic Functions for a Global Extremum Search in Solving Optimal Control Problems
RESUMO
The use of atomic functions in the theory of optimal control is shown, and the method for a global extremum search for a multi-parameter function with restrictions in the form of equalities and inequalities is described. To obtain local minimums a gradient method with a variable metrics is chosen. The method allows us to find a matrix of second derivatives for the function investigated at the point of a local minimum. This gives us the possibility to choose saddle-point directions and a step of movement along the directions in the domain of attraction of neighboring minimums. To exclude the domains that have been investigated when a local descent is realized, hyperspheres are constructed with radiuses equal to the iteration step. Results of experiments for the method on test problems are presented.