Task analysis of ultra-precision assembly processes for automation of human skills

A lot of research on assembly automation problems has been made by many researchers from a variety of communities, e.g., manufacturing, artificial intelligence and so on. They mostly focused on the geometric constraints related to a considered task, e.g., relative configuration change between the peg and the hole. Such an approach, however, seems valid only when the dimensional tolerance is fairly large so that it allows the relative configuration of the two mating parts to be changed in an obvious fashion. The paper investigates how highly precise assembly tasks can be automated. Tolerance considered in the present study is in the range of 3 /spl mu/m /spl sim/ 9 /spl mu/m. Currently the assembly task is being operated by skilled human workers. In order to analyze the insertion task, we focus on force sensory information and attitude measurements involved in the high-precision peg-in-hole task of a polygon mirror unit because a position-based analysis under such a narrow dimensional tolerance seems inappropriate. A variety of insertion experiments are conducted by skilled and unskilled subjects for comparison purposes.