Many products and components trend toward miniaturization for recent years. Micro parts will be more functional, so should be made of different materials, have a complex geometry, or is manufactured using incompatible processes. For those reason, the microassembly technology has become important and needed. In most microassembly tasks, manual assembly method or teleoperation method are used. Because, however, those methods are time consuming and highly expensive methods, automated method is positively necessary. In this thesis, for the automated and high precise microassembly, visual servoing techniques are used. Visual servoing techniques have shown great promise as a control strategy capable of micron precision while compensating for the many of the problems that exist in the micro domain, including imprecisely modeled and calibrated sensors and actuators.
Microassembly using visual servoing has some problems. For the high precise assembly, the high magnification lens should be used. It has two serious problems that are small field of view and small depth of field. The former problem is solved by using motorized zoom lens which can changes the field of view, and the latter problem is solved by using autofocusing method. Another problem in microassembly using visual servoing is occlusion of the feature. In this thesis we proposed the method that estimate the occluded feature by using the feature-tracking error. Using those methods, microassembly process separates into two processes that are macro visual servoing and micro visual servoing. Macro visual servoing is the process that makes an environment that can high precisely assemble the micro parts by zoom control and visual servoing. And micro visual servoing is the process that actually high precisely assembles the micro parts by visual servoing.
In this thesis, in order to proove the proposed algorithm on the real assembly task, the task of assembling the optical fiber and V-groove in the optical fiber array is experimented. And for the verification of visual servoing characteristics that is robust to the calibration errors, we experiment the case that the calibration errors exist.
