A purely vision-based trajectory description/control method is presented.
The concept of Augmented Image Space is introduced as a remedy for the non-linearity of a visual sensor. Essential image frames defined in this biologically motivated space can provide a general tool for describing robot's motion easily and flexibly.
To control the robot on-line, 3-D information extraction methods are investigated, and a special camera set called Coaxially aligned dual Camera Set is devised. A paractical software package which provides editing, compiling, and interactive executing capability is developed to organize the new concepts proposed in this paper.
Experiments using PUMA 560 robot manipulator are performed and numerous examples are illustrated to show the efficacy of the visual trajectory control scheme.