Abstract
This paper captures an assembly task as a process of transforming an initial contact state concept into that having only the goal state, where a contact state concept is the set of the possible contact states. When choosing a task element to be applied to a contact state concept, the operator has several options. The branch occurs at this point is named the decision branch. In addition, the operator can distinguish the transformed contact state concept into several types according to what she/he has sensed. The branch occurs at this point is called the observation branch. Thus, the whole strategy for a manual assembly task can be expressed with a tree having these branches, which is named the task sequence tree. The paper next discusses how to obtain a simple task sequence tree for a task a priori. Experiments confirm that teaching a simple task sequence tree of a task improves the efficiency of the task and also of further learning.
