On output regulation of direct visual servoing via velocity fields

Abstract

The velocity field approach has become very important in the last few years in robot control, since it has many advantages. For instance, it allows to describe a control objective for robot manipulators without providing a function of time. Despite this, on the contrary to usual point-to-point regulation, this method delivers a good transient response. Furthermore, most control algorithms assume that velocity measurements are available for feedback. For this reason, in this article it is shown that a well-known observer for robot manipulators can be used together with a visual servoing approach based on velocity fields. It is assumed that the robot is planar and the camera is fixed, so that the image plane is parallel to the manipulator workspace. Experimental results show the good performance of the complete system.

Keywords:

• visual servoing
• velocity fields
• tracking control

Acknowledgements

This work is based on research supported by the CONACYT under grant 58112 and by the DGAPA–UNAM under grants IN109306 and IN105408.

Notes

Notes

1. It should be noted that the controller tracks a trajectory in joint coordinates and not in image coordinates. After getting the desired final position for the robot end-effector, i.e. y _d, directly from the computer screen, a corresponding continuous trajectory in joint coordinates is obtained by using a velocity field, that is why the control-observer is given.

2. It is quite difficult to obtain a precise analytical model of a robot manipulator by direct computation of (42). However, we consider this to be an advantage since it allows to show the robustness properties of our approach.

3. The camera parameters can only be considered an approximation of the actual ones, since no standard calibration algorithm has been employed. For λ and α, data sheets have been used, was obtained by using a metric rule, φ = 0 comes from the camera position, u _o and v _o are gotten empirically by trial and error. As before, this allows to test the robustness of our approach.