Journal cover Journal topic
Mechanical Sciences An open-access journal for theoretical and applied mechanics
Journal topic

Journal metrics

IF value: 1.086
IF 5-year value: 1.369
IF 5-year
CiteScore value: 2.2
SNIP value: 0.87
IPP value: 1.38
SJR value: 0.306
Scimago H <br class='widget-line-break'>index value: 23
Scimago H
h5-index value: 15
Supported by
Logo Library of Delft University of Technology Logo NWO
Affiliated to
Logo iftomm
Articles | Volume 6, issue 2
Mech. Sci., 6, 191–201, 2015

Special issue: Modelling and control of robots

Mech. Sci., 6, 191–201, 2015

Research article 11 Sep 2015

Research article | 11 Sep 2015

A two-stage calibration method for industrial robots with joint and drive flexibilities

M. Neubauer1, H. Gattringer1, A. Müller1, A. Steinhauser1, and W. Höbarth2 M. Neubauer et al.
  • 1Institute of Robotics, Johannes Kepler University Linz, Altenbergestr. 69, 4040 Linz, Austria
  • 2Bernecker + Rainer Industrie Elektronik Ges.m.b.H., B & R Str. 1, 5142 Eggelsberg, Austria

Abstract. Dealing with robot calibration the neglection of joint and drive flexibilities limit the achievable positioning accuracy significantly. This problem is addressed in this paper. A two stage procedure is presented where elastic deflections are considered for the calculation of the geometric parameters. In the first stage, the unknown stiffness and damping parameters are identified. To this end the model based transfer functions of the linearized system are fitted to captured frequency responses of the real robot. The real frequency responses are determined by exciting the system with periodic multisine signals in the motor torques. In the second stage, the identified elasticity parameters in combination with the measurements of the motor positions are used to compute the real robot pose. On the basis of the estimated pose the geometric calibration is performed and the error between the estimated end-effector position and the real position measured with an external sensor (laser-tracker) is minimized. In the geometric model, joint offsets, axes misalignment, length errors and gear backlash are considered and identified. Experimental results are presented, where a maximum end-effector error (accuracy) of 0.32 mm and for 90 % of the poses a maximum error of 0.23 mm was determined (Stäubli TX90L).

Publications Copernicus