Institut für
Robotik und Prozessinformatik

Deutsch   English

Zur Zeit leider nur in Englisch vorhanden...

Friction Precompensation for Geared Robots

Project completed.

Project Description

In most industrial robots (IR), the torques of the motors are transformed by gears to the links (typically 50:1...200:1). The dynamical behaviour of robots is significantly affected by gears; a high amount of torque (in general: force or torque) is needed to overcome friction. In contrast to direct driven robots, the frictional torque is typically in the same range as the torque needed to move or to hold the robot link itself. A detailed knowledge of friction is essential for any research in robot dynamics, especially for geared robots. With the knowledge about the frictional behaviour, a precompensation can be applied. Friction precompensation releases the controller from the compensation of the disturbances of nonlinear frictional effects and contributes to a more accurate robot control.

In literature, frictional effects are mainly classified with respect to joint velocity. The following effects are distinguished: static friction, viscous friction, stiction, breakaway, slip stick, and negative friction. Although some friction models, which are known from literature, are very detailed and sophisticated, temperature dependencies have not been considered so far. 

Decrease of velocity-dependent friction effects with increasing temperature.

As shown by the plot above, the velocity dependent frictional effect decreases with increasing temperature. This can be explained by the viscousness of the lubricant in the bearings and in the gears. In addition we have measured friction at very low speeds and thus were able to identify negative viscous friction.
Even position dependent friction can be observed, which is related to the teeth of the gears and thus also is periodic with the motor revolution. This friction component can be seen at very low speeds only; otherwise the inertia of the link itself damps this friction ripple. Hence it has not been taken into account in our friction model.
In order to obtain a model for the measured behaviour, the function shown in the plot above has been approximated with orthogonal polynomials. The basis functions have been adapted due to the knowledge from experiments, i.e., friction dependency is almost linear with the temperature.
Often, the torques imposed on the joints are assumed to be proportional to the applied motor currents. But, as can be seen from above , up to 1/3 of the maximum motor current is necessary to overcome frictional effects. With full friction compensation the links can be considered as frictionless, i.e., we have the possibility to directly apply desired torques to the links.

Structure of the precompensation approach.

The left plot shows velocity measurements of a robot joint without friction precompensation; the curve parameters are different constant motor current values. As can be seen, the velocity reaches a near constant value after approx. 200ms; this corresponds to a balance between motor torque and friction torque. Using the above outlined precompensation scheme, the velocity is significantly linearized, as shown in the right plot. Even at very small desired values (10 or 100) the joint is accelerated constantly, i.e., velocities are increasing linearly. Note that without friction precompensation the output value to overcome static friction was about 2500 inc.As test environment we used a 6 dof IR manutec r15.

Joint velocity without friction precompensation.

Joint velocity with friction precompensation.

It took 0.29s to generate this page.