Identification and Compensation of Gear Friction for Modeling of Robots
The identification of the dynamics of a standard industrial robot is solved by the application of the multivariable least square method. In order to eliminate the dominant influence of friction in gears and joints a nonlinear friction model is adapted to
- PDF / 1,131,954 Bytes
- 8 Pages / 481.89 x 691.654 pts Page_size
- 37 Downloads / 239 Views
M. Daemi and B. Heimann University of Hannover, Hannover, Germany
Abstract - The identification of the dynamics of a standard industrial robot is solved by the application of the multivariable least square method. In order to eliminate the dominant influence of friction in gears and joints a nonlinear friction model is adapted to measured friction characteristics. Its influence is compensated in the identification step. The base parameter vector is grouped and optimal trajectories are used to identify each group. The quality of the identified model is verified by comparison of measured trajectories and torques predicted by the model.
Introduction A common approach in robotics for the identification of the equation of motion is the use of multivariable least square methods. It allows a relatively fast identification of a set of linear base parameters of the robot without the need of disassembling it. A main concern in practical applications of this method is the large influence of the friction in gears and joints and their variation with changing operating conditions. In this paper the identification of the 6-dof dynamic model of a standard industrial robot is presented, which emphasizes on the prediction of a precise friction model and the compensation of its influence on the identification process. The measurements rely only on the robots internal incremental encoders and current sensors for torque measurements. Although this work is adapted to the characteristics of the manutec-r15, the approach can be used for most industrial robots. A common method for smaller systems (dof s; 3) to avoid the influence of the variation of friction parameters is to include them into the identification process by modeling a term of dry friction and a term for viscous damping (e.g.[!]). This approach would enlarge the parameter vector for a 6-dof robot by another 12 elements and would lead to problems in the identification process. In this paper effects of the friction model for each link are evaluated separately and their influence on the measured torques are compensated before starting the identification. In order to enable the excitation of all parameters of the dynamic model, an optimization scheme is proposed for defining trajectories that yield good identification results.
Modeling of gear friction The torques exerted by the drives of each joint of the robot consist of a number of effects which are: acceleration of the moments of inertia, centrifugal- and Coriolis forces, A. Morecki et al. (eds.), ROMANSY 11 © Springer-Verlag Wien 1997
90
M Daemi, B. Heimann !
1
... i
-···
150
0.5
I
o
Data Loading...
