Guiding a robot by visual feedback in assembling tasks

INTELLIGENT robots which manipulate objects using visual information have been studied and experimental models have already been developed in various laboratories. ~1-3) Generally the intelligent robot consists of the visual system, the handling system and the supervisory system that links these systems. The visual system and handling system are usually combined together as an open loop system. Once the visual information is processed and the parameters are obtained, the visual system is completely separated from the handling system, and while the objects are handled, the visual system is not employed. In this open loop system, the accuracy of the operation depends directly on the errors of both the visual input device and the digitization in the visual system, the positioning errors in the handling system and the errors in coordinate transformation. In order to compensate for those errors, a visual feedback loop can detect the difference between the actual position and the desired position of the manipulator. If the manipulator's position is adjusted to the desired position by the visual feedback loop, the accuracy of the task is raised to the point of the resolution of both the visual and handling systems. In this paper, the assembly task of placing a square prism block into a square box is described as an example of an operation with a visual feedback loop. In this operation, the recognition problem is easy: however, it requires the accurate control of position and orientation.