Primary objective of this research is to thoroughly analyze and model a high-precision assembly task which is a peg insertion with a very narrow tolerance, i.e., 5 micron or less. Currently the task has been being performed by skilled workers because there is no robot or assembly machine available at this moment that can accomplish the task with a sufficient accuracy and efficiency that the skilled workers can deliver. A primary objective of the present study is to conduct in-depth analysis of the insertion task based on the experimental observations using high precision instruments. In doing so, we focused on collecting and analyzing the physical data associated with the insertion task in view of automatization of the task.
During the first year, we measured reaction forces with a commercially available force sensor and did comparisons between skilled workers and unskilled workers. There were a lot of distinctions observed between the skilled and the unskilled in terms of the force pat
terns and the magnitude of the forces they exerted. However, the force sensor used was not sensitive enough to give us all the information we needed, i.e., the resolution of moments was so low that any meaningful information could not be extracted. In addition to the force-related measurements, the inclination of the polygon mirror is measured as well.
In the second year of the research, we further conducted experiments after newly designing a 2-axis moment sensor which measures moments about X-axis and Y-axis. Its sensitivity is high enough whereas its cross talk as well as nonlinear properties are sufficiently low. Also a real-time image processing is used to obtain more accurate measurement of the inclination of the insertion jig. As a result, we found that not only skilled subjects can finish the task faster than the unskilled, but also the former tends to minimize unnecessary forces while the latter rather poorly searches for an incidental insertion. This observation is supported by the fact that the inclination change of a skilled subject more or less directly converge to a goal while an unskilled subject tends to do a processing motion in order to find an exact alignment, which in general takes more time and is less energy efficient. The way the skilled subjects approach the task is highly consistent and efficient. Such consistency and efficiency must be related with tactile sensitivity which they possess. Less