Complexity of neurocontrol algorithms

Author

Hrycej, Tclmas

Author_Institution

Res. Center, Daimler-Benz AG, Ulm, Germany

Volume

2

fYear

1998

fDate

4-9 May 1998

Firstpage

949

Abstract

Critic-based approaches and closed-loop optimization are two of the most important fundamental neurocontrol approaches, which can be used in incremental or batch mode. To assess their complexity for the same type of nonlinear control problems, idealized algorithms on a discrete state space are constructed, using a dynamic optimization framework. The comparison of both algorithm complexity and number of data samples necessary to reach the solution is done. Alternative complexity investigation is done by comparing the number of parameters to be instantiated in linear case. The incremental processing turns out to be the less efficient alternative

Keywords

Lyapunov methods; computational complexity; discrete systems; dynamic programming; neurocontrollers; nonlinear control systems; optimal control; state-space methods; batch mode; closed-loop optimization; critic-based approaches; discrete state space; dynamic optimization framework; idealized algorithms; incremental mode; neurocontrol algorithms; nonlinear control problems; Backpropagation; Computer networks; Convergence; Cost function; Neural networks; Neurocontrollers; Optimal control; Optimization methods; Sampling methods; State-space methods;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks Proceedings, 1998. IEEE World Congress on Computational Intelligence. The 1998 IEEE International Joint Conference on

Conference_Location

Anchorage, AK

ISSN

1098-7576

Print_ISBN

0-7803-4859-1

Type

conf

DOI

10.1109/IJCNN.1998.685898

Filename

685898