7 Conclusion

An important component of the system presented in this paper has been affine stereo. The appendix contains a quantitative comparison with perspective stereo. Affine stereo provides a simple and robust interpretation of image position and disparity that degrades gracefully when cameras are disturbed. Calibration is not only easier (fewer parameters and amenable to linear techniques) but also less sensitive to small perspective effects. It is suitable for uncalibrated and self-calibrating systems and therefore the preferred stereo method for our visual servoing application.

By defining the working coordinate system in terms of the robot's abilities, linear errors in its kinematics are bypassed. The remaining nonlinearities can be handled using image-based control and feedback. We have shown that this can be achieved cheaply and effectively using a novel form of active contour to track planar features on the gripper and target.

Such a system has been implemented and found to be highly robust, without unduly sacrificing performance (in terms of speed to converge on the target).


  • Appendix
  • References
  • Back to Contents