An approach to hand/eye coordination that does not require any calibration and hence overcomes some of the shortcomings of traditional approaches is presented. It is demonstrated how the concepts of the perceptual kinematic map and control surface of the hand/eye system allow the modeling and solution of the grasping problem. The close integration of visual information with the decision process and reliance on qualitative, topological information rather than on the exact values of the maps allow robust visual control of the robot manipulator. The principles of the approach are general, and the algorithm and system presented here represent only one of many possible ways to implement and test the approach.
展开▼