Scientific journal
Scientific Review. Technical science
ISSN 2500-0799
ПИ №ФС77-57440

Personal portfolio

SYMBOLIC REGRESSION METHOD BASED ON NETWORK OPERATOR TO SOLVE THE PROBLEM OF CONTROL SYNTHESIS

Shmalko E.Y. 1 Diveev A.I. 1
1 Institution of Russian Academy of Science Dorodnicyn Computing Centre of RAS
The paper concentrates on the problem of control synthesis that is to find a multi-dimentional control function which depends on the coordinates of the state space. Such problem to search a needed mathematical expression relates to symbolic regression tasks. Nowadays it becomes possible to use a computer for symbolic regression calculations. A symbolic regression method based on the network operator is proposed in order to automate the search of a control function. The network operator is a special data structure that allows to present any mathematical expression in the form of a matrix with numbers of unary and binary operations as elements. Changing the elements of the matrix allows to receive new mathematical expressions. Such approach makes possible the usage of various search algorithms to find an optimal mathematical expression. The intellectual evolution algorithm is applied for searching. The paper provides a numerical example of path motion of mobile robot with constraints of state.