A BPTT-like Min-Max Optimal Control Algorithm for Nonlinear Systems (CROSBI ID 566206)
Prilog sa skupa u zborniku | izvorni znanstveni rad | međunarodna recenzija
Podaci o odgovornosti
Milić, Vladimir ; Kasać, Josip ; Majetić, Dubravko ; Šitum, Željko
engleski
A BPTT-like Min-Max Optimal Control Algorithm for Nonlinear Systems
This paper presents a conjugate gradient-based algorithm for feedback min-max optimal control of nonlinear systems. The algorithm has a backward- in-time recurrent structure similar to the back propagation through time (BPTT) algorithm. The control law is given as the output of the one- layer neural network. Main contribution of the paper includes the integration of BPTT techniques, conjugate gradient methods, Adams method for solving ODEs and automatic differentiation (AD), to provide an effective, novel algorithm for solving numerically optimally min-max control problems. The proposed algorithm is applied to the rotational/translational actuator (RTAC) nonlinear benchmark problem with control and state vector constraints.
Min-Max optimal control; Conjugate gradient methods; Back-propagation through time; Automatic differentiation; RTAC system
nije evidentirano
nije evidentirano
nije evidentirano
nije evidentirano
nije evidentirano
nije evidentirano
Podaci o prilogu
539-542.
2010.
objavljeno
Podaci o matičnoj publikaciji
Numerical Analysis and Applied Mathematics
Theodore E. Simos, George Psihoyios, Ch. Tsitouras
Melville (NY): American Institute of Physics (AIP)
978-0-7354-0831-9
Podaci o skupu
ICNAAM 2010: International Conference of Numerical Analysis and Applied Mathematics 2010
predavanje
19.09.2010-25.09.2010
Rodos, Grčka