Investigation of the effect of time scaling in a soft computing based control using fractional order derivatives

Tar, József K.; Rudas, Imre J.; Bitó, János F.; Machado, J. A. Tenreiro; Szeghegyi, Ágnes

http://hdl.handle.net/10400.22/9761

Use this identifier to reference this record.

Name:	Description:	Size:	Format:
COM_MachadoTenreiro169_2004.pdf		863.69 KB	Adobe PDF	Download

Send Feedback

Authors

Tar, József K.

Rudas, Imre J.

Bitó, János F.

Machado, J. A. Tenreiro

Szeghegyi, Ágnes

Abstract(s)

In this paper the time-scaling properties of an adaptive control based on a novel branch of Computational Cybernetics is investigated in the control of an approximately modeled non-linear system. Adaptive learning of the system's dynamic properties makes it possible to prescribe the desired tracking error relaxation by purely kinematic terms also containing fractional order time-derivatives for which Caputo's definition is used. This definition leads to nonlinear dependence in the lenght of the cycle time of the controller numerically estimating the time-derivatives. Therefore the division of the cycle-time of the central control unit responsible for dealing with dynamic coupling to smaller intervals available for the local controllers influences the quality of the control as a whole. Simulation results are given to demonstrate the limits of time-resolution that can be applicable in the case of this approach. It is concluded that the method is relatively robust, and that the most important limit is set to the cycle-time of the central unit.