Learning control for accuracy enhancement of parallel kinematic machines

verfasst von

H. Abdellatif, Bodo Heimann

Abstract

The aim of this research is to qualify parallel robots for high-precision dynamic tasks using appropriate control strategies. Therefore, the prototype was designed for a convenient weight-to-load-ratio and is equipped with electromagnetic linear direct drives. It has been found that despite the feedforward compensation of nonlinear dynamics and the accurate identification of its parameters, some tracking errors remain. This is due to non-modeled effects and primarily permanent magnetic and electromagnetic actuator disturbances. To further increase the control quality ILC (iterative learning control) was applied to the parallel highly coupled mechanism. In addition, it was questionable that direct drives allow simple formal decoupling of the complex MIMO-system into multiple SISO-systems. Direct drives could be demonstrated to be acceptable only in case of conventional industrial robots or mechanisms, where high gear ratios are used. A phase-lead compensated approach which can be easily parameterized by experimental investigation of the system's frequency response was chosen. Improvement of the tracking performance was demonstrated in time- and frequency-domains as well as in the actuation and in the operational space. Independently of the actuator or the motion, ILC keeps a good performance after the low number of 15 iterations, which is good argument for its practicability. For repetitive operations, it is possible to keep the ILC as a task in the background that calculates corrections for the input of every following operation.

Organisationseinheit(en)

Institut für Mechatronische Systeme

Typ

Aufsatz in Konferenzband

Seiten

443-456

Publikationsdatum

2006

Publikationsstatus

Veröffentlicht

Peer-reviewed

Ja